null  null
-1-
IMPROVING HIGH-FREQUENCY AUDIBILITY
for
HEARING-IMPAIRED LISTENERS
using a
COCHLEAR IMPLANT
or
FREQUENCY-COMPRESSION AID
Andrea Simpson
Submitted in total fulfilment of the requirements for the degree of
Doctor of Philosophy
Department of Otolaryngology
Faculty of Medicine, Dentistry and Health Sciences
The University of Melbourne
(May 2007)
-2–
-3-
Abstract
Three experiments were carried out on adult hearing-impaired listeners with
sloping-sensorineural losses to determine an optimum method for providing
high-frequency information. For the first experiment, a consonant identification
test was carried out with 10 listeners with moderate-to-profound high-frequency
hearing losses under various low-pass filter conditions. Subjects were also
tested for cochlear dead regions with the TEN test. Consonant recognition was
tested under conditions in which the speech signals were highly audible to
subjects for frequencies up to the low-pass filter cutoff. Extensive dead regions
were found for one subject with the TEN test. Average consonant scores for the
subject group improved significantly with increasing audibility of high-frequency
components of the speech signal. There were no cases of speech perception
being reduced with increasing bandwidth. Nine of the subjects showed
improvements in scores with increasing audibility, whereas the remaining
subject showed little change in scores. For this subject, speech perception
results were consistent with the TEN test findings. In general, the results
suggest that listeners with severe high-frequency losses are often able to make
some use of high-frequency speech cues if these cues can be made audible.
For the second experiment, an experimental wearable frequency-compression
hearing aid was selected to investigate whether the device could improve
speech perception in a group of 7 listeners with steeply-sloping audiograms. In
the frequency-compression hearing aid, high frequencies (above 1600 Hz) were
amplified in addition to being lowered in frequency. Lower frequencies were
amplified without frequency shifting. No significant differences in group mean
scores were found between the frequency-compression device and a
conventional hearing device for understanding speech in quiet. Testing in noise
showed improvements for the frequency-compression scheme for only one of
the five subjects tested. Subjectively, all but one of the subjects preferred the
sound quality of the conventional hearing instruments. In conclusion, the
experimental frequency-compression scheme provided only limited benefit for
these listeners.
-4–
For the third experiment, 5 subjects with steeply-sloping losses underwent
cochlear implantation. The surgery technique was modified in an attempt to
preserve subjects’ residual hearing. An in-the-ear (ITE) hearing aid together
with the speech processor was fitted to subjects post-operatively if useful
hearing was present in the implanted ear. Psychoacoustic measures were used
to create a map in which the implant was programmed to provide the listener
with high-frequency information above the frequency at which acoustic hearing
was no longer considered useful. Speech recognition testing in quiet and noise
with this map were compared with a second map, which provided the fullfrequency range.
Four of the five subjects had some measurable hearing thresholds postoperatively. Two subjects were fitted with an ITE hearing aid post-operatively.
For the group, cochlear implantation provided superior recognition of
monosyllabic words, consonants, and sentences in noise compared to preoperative scores. For testing in quiet and noise, there were no significant
differences between the two maps when subjects were wearing their implant
together with hearing aid/s. In conclusion, for this study, the combination of
cochlear implants together with hearing aids was the most effective at providing
an alternative rehabilitation for hearing-impaired listeners with steeply-sloping
losses.
-5Declaration
This is to certify that
(i)
the thesis comprises only my original work towards the PhD except
where indicated in the preface,
(ii)
due acknowledgement has been made in the test to all other material
used,
(iii)
the thesis is less than 100,000 words in length, exclusive of tables,
maps, bibliographies and appendices.
-6–
Preface
I am grateful for the financial support of the Commonwealth of Australia through
the Cooperative Research Centres program.
Phonak Hearing Systems provided the hearing devices used in the study.
Cochlear Ltd provided the Nucleus Freedom systems used in 5 of the subjects.
Several subjects were recruited from the Melbourne Cochlear Implant Clinic.
I would like to acknowledge the many colleagues who contributed to this work.
The frequency-compression device described in Experiment 2 was initially
investigated in a study carried out together with Adam Hersbach and Hugh
McDermott. Input into building the devices was provided by Justin Zakis. Many
of the software programs used in the current study have been designed and
maintained by the following people: Hugh McDermott, Justin Zakis, Adam
Hersbach, and Rodney Millard. John Heaseman provided technical support for
the CI recipients in this study. He also helped to create the individual mapping
programs for each patient. Waikong Lai created the MACarena software
package used to carry out speech testing.
Cathy Sucher carried out some of the conventional hearing-aid testing
described in Experiment 3 of this study.
Advice was sought from Graham Hepworth at the Statistical Consulting Centre
at the University of Melbourne.
There are many other individuals who were helpful in setting up the project.
These were Lesley Whitford, Peter Busby, Kerrie Plant, Liz Winton, Sylvia Tari,
and Leonie Fewster. Robert Briggs and Stephen ‘O Leary performed the
surgery required for Experiment 3. Harvey Dillon, Brian Moore, Dave Fabry,
Silvia Allegro, and Susan Scollie all provided valuable advice about the nature
of the project.
Finally, the study would not have been possible without the subjects who
generously gave of their time to participate in the experiments.
-7-
List of Publications
Simpson, A., Hersbach, A.A., McDermott, H.J., 2005a. Improvements in
speech perception with an experimental nonlinear frequency-compression
hearing device. Int J Audiol., 44, 281-292.
Simpson, A., Hersbach, A.A., McDermott, H.J., 2005b. Relationship
between speech recognition and self-report measures. Acoustics Australia, 33,
57-62.
Simpson, A., McDermott, H.J., Dowell, R.C., 2005c. Benefits of audibility
for listeners with severe high-frequency hearing loss. Hearing Research, 210,
42-52.
Simpson, A., Hersbach, A.A., McDermott, H.J., 2006. Frequency
compression outcomes for listeners with steeply sloping audiograms. Int J
Audiol., 45, 619 - 629.
Abstracts
McDermott H.J., Simpson A., Sucher C.M. “Acoustic-electric stimulation
with the Nucleus Freedom system,” 9th International Conference on Cochlear
Implants, Hofburg Emperor Palace, Vienna, Austria, 14-17 June 2006.
Simpson, A., McDermott, H.J., Dowell, R.; Sucher, C.; Cowan, R.
“Combining cochlear implants and hearing aids together for listeners with
steeply sloping losses.” abstracts of the XVII National Conference of the
Audiological Society of Australia 2006, p 43, Perth, Australia, Australian and
New Zealand Journal of Audiology, vol. 28, supplement 2006.
Sucher, C.; McDermott, H.J.; Simpson, A. “Comparing the perceived
pitch of acoustically and electrically presented stimuli.” abstracts of the XVII
National Conference of the Audiological Society of Australia 2006, p 45, Perth,
Australia, Australian and New Zealand Journal of Audiology, vol. 28,
supplement 2006.
-8–
McDermott H.J., Simpson A., Sucher C.M. “Recent advances in
compensating for severe high-frequency hearing impairment,” 21st Danavox
Symposium on Hearing Aid Fitting, Kolding, Denmark, 31 August - 2 September
2005 (invited)
McDermott H., Simpson A., Hersbach A. “Frequency compression can
enhance speech perception for hearing-aid users,” abstracts of the International
Hearing Aid Research Conference IHCON 2004, p 25, Lake Tahoe, California,
USA, 25-29 August 2004. (invited).
Simpson A., Hersbach A.A., McDermott H.J. “Performance of an
experimental frequency-compression hearing aid,” abstracts of the XVI National
Conference of the Audiological Society of Australia (May 17 - 20, 2004),
Brisbane, Australian and New Zealand Journal of Audiology, vol. 26,
Supplement 2004, pp. 40 - 41.
Simpson A., Hersbach A.A., McDermott H.J. “Performance of a Novel
Frequency-Compression Hearing Aid,” abstract and handout PP126, American
Academy of Audiology conference (31 March – 3 April 2004).
-9Acknowledgements
A paragraph consisting of a few sentences does not give justice to the support
both my supervisors have provided me with over the past 3 years. I consider
myself very lucky to have worked closely with Professor Hugh McDermott,
initially as a research assistant and later as his PhD student. His supervision
has fostered in me a love for research and this work would not have been
possible without his astounding mentorship abilities. I cannot overemphasize
how much he has taught me about conducting research. I have to thank him for
sharing his knowledge, ideas, and creativity so freely. At times, I experienced
this project to be a particularly emotionally harrowing experience. I confess to
frustrated, bad-tempered days when I was not especially open to the idea of
constructive criticism. His patience in dealing with me on a personal as well as
profession level is much appreciated. I’d especially like to recognize the many
hours he has put into making this project a success.
This work would not have run nearly as smoothly without the help of my second
supervisor, Professor Richard Dowell. In fact, without him I may still have
been in the process of trying to convince several parties of the validity of the
work. His enthusiasm, belief in me, and excellent diplomacy skills “saved the
day” on more than one occasion. Perhaps most importantly, both Richard and
Hugh allowed me independence and space to grow. I always felt as though I
was taken seriously as a colleague rather than a student and for this they both
have my gratitude and respect.
My thanks go to Associate Professor Robert Cowan and Dr Stefan Launer
who were extremely supportive of the project and helped to create my
scholarship as well as other funding through the CRC for Cochlear Implant and
Hearing Aid Innovation. Both have also commented on previous versions of my
publications.
Adam Hersbach has been a fantastic person to work with and has provided me
with much engineering technical support and knowledge. I also highly
recommend him as a person to share a beer over or to take a stroll around San
Francisco with.
- 10 –
Cathy Sucher is an amazing audiologist and friend who somehow manages to
successfully juggle family together with work and study. She also found some
time to carry out some of the testing in this project and was always helpful and
supportive.
Finally, to my family and friends. My loving family in South Africa, especially my
mom who endured several tearful phone conversations. I have made some
wonderful friends and they all deserve a book dedicated in thanks to them.
Three in particular deserve a mention. All successful, driven, smart woman who
have been through or are about to embark on the PhD adventure themselves.
Thank you, Vanessa Surowiecki and Jacqueline Andrew for your humour and
sympathetic shoulders. Finally, to Miss Carrie Newbold, who has provided me
with so much emotional support. Even at times when I knew she was facing
some pretty big stresses herself. Thank you for the times of letting me rant,
whine, cry, rage, over-analyze, and over-sensitize all the issues in my life
without ever judging me or closing your door.
-11-
Table of Contents
Chapter 1......................................................................................................... 17
1.1
Aims and Objectives .............................................................................. 22
Chapter 2……………………………………………………………………………..25
2.1
Introduction............................................................................................ 25
2.2
Anatomy of the inner ear ....................................................................... 25
2.3
Perception of sound in the auditory system ........................................... 27
2.3.1 Cochlear tuning and frequency selectivity ...................................... 29
Chapter 3......................................................................................................... 35
3.1
Introduction............................................................................................ 35
3.2
The Speech Intelligibility Index (SII)....................................................... 35
3.3
Audibility studies with hearing-impaired listeners .................................. 36
3.4
The definition of a dead region .............................................................. 40
3.4.1 Effects of off-frequency listening..................................................... 41
3.5
Methods of detecting dead regions........................................................ 42
3.5.1 Psychophysical tuning curves ........................................................ 42
3.5.2 Masking with Threshold-Equalizing Noise (TEN)............................ 43
Chapter 4......................................................................................................... 49
4.1
Introduction............................................................................................ 49
4.2
Early attempts at frequency shifting....................................................... 49
4.3
Frequency shifting by means of a “slow-play” method........................... 52
4.4
Speech perception results with a novel frequency-shifting device......... 53
Chapter 5......................................................................................................... 59
5.1
Introduction............................................................................................ 59
5.2
Cochlear implant system design ............................................................ 59
5.3
Current implant systems commercially available ................................... 62
5.4
Speech-coding strategies ...................................................................... 64
5.4.1 The Advanced Combination Encoder (ACE) speech-processing
strategy ..................................................................................................... 67
Chapter 6......................................................................................................... 73
6.1
Introduction............................................................................................ 73
6.2
Electrode insertion depth and frequency-place mismatch in the
cochlea………………………………………………………………………...73
6.3
Combining electric and acoustic stimulation .......................................... 81
6.3.1 Combining electric and acoustic stimulation in opposite ears......... 82
6.3.2 Combining electric and acoustic stimulation in the same ear ......... 84
Chapter 7......................................................................................................... 95
7.1
Introduction............................................................................................ 95
7.2
Subjects................................................................................................. 95
7.2.1 Criteria for recruitment.................................................................... 95
7.2.2 Ethics approval ............................................................................... 96
- 12 –
7.3
Methods and procedures ....................................................................... 98
7.3.1 Introduction..................................................................................... 98
7.3.2 Measurement of hearing thresholds ............................................... 98
7.3.3 Measurement of dead regions ........................................................ 98
7.3.4 Measurement of aided thresholds ................................................ 101
7.3.5 Speech recognition testing ........................................................... 109
7.4
Results................................................................................................. 111
7.5
Discussion ........................................................................................... 117
7.5.1 Conclusion.................................................................................... 120
Chapter 8....................................................................................................... 121
8.1
Introduction.......................................................................................... 121
8.2
Methods and Procedures..................................................................... 121
8.2.1 Subjects........................................................................................ 122
8.2.2 Own-aid testing............................................................................. 126
8.2.3 Conventional device fitting and testing ......................................... 127
8.2.4 Frequency-compression fitting and testing ................................... 131
8.3
Results: own aid versus conventional device ...................................... 134
8.3.1 Word recognition in quiet.............................................................. 135
8.4
Results: conventional device versus frequency compression.............. 137
8.4.1 Word recognition in quiet.............................................................. 137
8.4.2 Consonant recognition in quiet ..................................................... 138
8.4.3 Sentence recognition in noise ...................................................... 140
8.4.4 Subjective assessment................................................................. 142
8.5
Discussion ........................................................................................... 143
8.5.1 Conclusions .................................................................................. 148
Chapter 9....................................................................................................... 149
9.1
Introduction.......................................................................................... 149
9.2
Methods and Procedures..................................................................... 149
9.2.1 Subjects........................................................................................ 149
9.2.2 Surgical technique ........................................................................ 151
9.2.3 Pitch estimation testing................................................................. 152
9.2.4 Pre-operative fitting procedure ..................................................... 155
9.2.5 Post-operative fitting procedure.................................................... 160
9.2.6 Monitoring of hearing thresholds .................................................. 164
9.2.7 Speech recognition testing ........................................................... 164
9.2.8 Subjective assessment................................................................. 164
9.3
Results................................................................................................. 165
9.3.1 Pre- and post-operative audiograms ............................................ 166
9.3.2 Pitch estimation results................................................................. 168
9.3.3 Speech perception results ............................................................ 170
9.3.4 Subjective assessment................................................................. 189
9.4
Discussion ........................................................................................... 191
9.4.1 Conclusions .................................................................................. 204
Chapter 10..................................................................................................... 205
Future research .......................................................................................... 208
-13-
- 14 –
List of Figures
Figure 2.1 Cross- sectional view of the human ear. ……………………………..34
Figure 2.2 Cross section of the Organ of Corti as it appears in the basal turn of
the cochlea.….………………………………………………………………..34
Figure 2.3 Representation of the way acoustic signals are passed from the
basilar membrane to the auditory nerve…………………………………...36
Figure 2.4 Schematic representation of the tuning of an auditory nerve fiber…39
Figure 2.5 Schematic representation of an auditory-nerve tuning curve……….40
Figure 3.1 Schematic representation of PTC results……………………………..51
Figure 4.1 An example of the input-frequency to output-frequency
relationship implemented in the experimental
frequency-compression hearing device………………………………...….62
Figure 4.2 Mean phoneme, consonant, fricative, and vowel scores
obtained by the 17 hearing-impaired subjects who participated in
the study when listening to monosyllabic words………………………….64
Figure 5.1 A cochlear implant system……………………………….……………..68
Figure 5.2 Representation of the three types of electrode configurations
used in cochlear implant systems………..………………………………...69
Figure 5.3 The Nucleus Freedom CI system…………………………………….. 71
Figure 5.4 The two types of stimulation waveform used in implant systems…..74
Figure 5.5 A representation of the ACE speech-processing strategy…………..76
Figure 5.6 Frequency allocation for the current default clinical filter bank
used in Melbourne, Australia with the Freedom device when the
ACE strategy is selected…...………………………………………………..77
Figure 5.7 An example of the amplitude conversion function for an
IDR of 45 dB. ………………………..……………………………………….78
Figure 6.1 Block diagram representing signal processing in a noise
Vocoder simulation…………………………………………………………..83
Figure 6.2 Schematic diagram representing the effects of
frequency-place expansion and compression investigated by
Baskent & Shannon (2003)………………………………………………….86
Figure 7.1 Air-conduction thresholds in dB HL for all subjects. Arrows
indicate no response at the audiometer’s maximum output…….……..105
Figure 7.2 Spectrum of the TEN for 70 dB/ERB (Figure reproduced
from Moore, 2001)…………….……………………………………………108
Figure 7.3 Frequency response of the Supero 412 hearing instrument
in response to the MACarena broadband calibration noise at a
level of 60 dBA
…..……………………………………………………112
Figure 7.4 Screen shot of the MACarena test format…………………………..119
Figure 7.5 TEN test results for S01-S10…………………………………………120
Figure 7.6 Consonant test mean scores (filled circles) for S01-S10 are
indicated on the right y-axis as percentage correct………………..…...123
Figure 7.7 Average information transmission scores. ………………………….127
Figure 8.1 Hearing threshold levels (dB HL) for the subjects who
participated in Experiment 2. ………………………………….………….132
Figure 8.2 Flow chart of the testing carried out for Experiment 2 for
each of the 7 subjects who participated in the study…………..……….133
Figure 8.3 Block diagram of the binaural signal processing
implemented in the frequency compression hearing device. ….………139
-15Figure 8.4 Mean phoneme scores for the left and right ears when
listening to monosyllabic words obtained by the experienced
hearing-aid users who participated in the study. …………………….....144
Figure 8.5 Mean phoneme scores for the left and right ears when
listening to monosyllabic words obtained by the non-hearing-aid
users who participated in the study…..…………………………………..145
Figure 8.6 Mean phoneme scores for the binaural, left-ear alone,
and right-ear alone conditions for the 7 subjects who participated
in Experiment 2 when listening to monosyllabic words. ……………….146
Figure 8.7 Mean consonant scores obtained by the 7 hearing-impaired
subjects who participated in Experiment 2…….…………………………148
Figure 8.8 Mean signal-to-noise ratios (SNRs) obtained by the subjects…….149
Figure 8.9 Preference scores from the APHAB questionnaire provided
by 6 subjects who participated in the study. ………...…………………..150
Figure 9.1 Flow chart of the testing carried out for Experiment 3 for
each of the 5 subjects who participated in the study…..……………….158
Figure 9.2 Schematic representation of the frequency allocations for
the CI and both hearing aids for the place-matched map for
the 5 subjects who participated in Experiment 3………………………..167
Figure 9.3 Pre- and post-operative audiogram results for the 5
subjects who participated in the study. …………………………………..175
Figure 9.4 Pitch estimation results measured at 2 weeks postoperatively for the 5 subjects who participated in Experiment 3. ……..177
Figure 9.5 Phoneme scores for the CNC word test for the 5 subjects
who participated in Experiment 3. ………………………………………..182
Figure 9.6 Consonant scores for the CNC word test for the 5 subjects
who participated in Experiment 3. ………………………………………..184
Figure 9.7 Fricative scores for the CNC word test for the 5 subjects
who participated in Experiment 3. ………………………………………..186
Figure 9.8 Vowel scores for the CNC word test for the 5 subjects
who participated in Experiment 3. ………………………………………..188
Figure 9.9 Results for the consonant test for the 5 subjects who
participated in Experiment 3. ……….…………………………………….191
Figure 9.10 Mean signal-to-noise ratios (SNRs) obtained by the 5
subjects who participated in Experiment 3. ……………………………..195
Figure 9.11 Preference scores from the APHAB questionnaire
at 12 weeks post-operatively provided by the 5 subjects who
participated in the study…………………………………………………....198
Figure 9.12 Preference scores from the APHAB questionnaire carried
out after 28 weeks post-operatively provided by the 5 subjects
who participated in Experiment 3. ……….……………………………….199
Figure 9.13 Monosyllabic CNC word scores for Hybrid patients………………203
- 16 –
List of Tables
Table 5.1 Summary of published studies reporting monosyllabic word
recognition using current CI speech-processing strategies.. ..................... 67
Table 6.1 Summary of studies investigating the benefits of wearing a CI
together with a HA in children and adults.. ................................................ 85
Table 6.2 Success of hearing preservation in recently reported studies .......... 87
Table 7.1 Relevant information regarding the subjects who participated in
Experiment 1. ............................................................................................ 99
Table 7.2 Recorded output levels of the Supero 412 via an ear simulator,
for input levels corresponding to the LTASS (Byrne et al., 1994)............ 108
Table 7.3 Table used to determine the amount of gain necessary to
achieve aided thresholds 18 dB below each one-third octave band
level of the LTASS (Byrne et al., 1994). .................................................. 109
Table 7.4 Results of the Analysis of Variance for the consonant recognition
test. Asterisk symbols indicate interactions among factors. .................... 118
Table 7.5 Three-way table of means of subjects’ results................................ 119
Table 8.1 Relevant information about the subjects (S32 – S39) who
participated in Experiment 2, and their hearing aids. .............................. 125
Table 8.2 Confusion matrices for the consonant test. Correct responses
for each phoneme are shown in bold type diagonally. ............................ 147
Table 9.1 Frequency-to-electrode allocations in the current clinical
default map, used in the Melbourne Cochlear Implant Clinic,
assuming all 22 electrodes are active. .................................................... 160
Table 9.2 Speech recognition and subjective test conditions included
for the 5 subjects who participated in the study....................................... 167
Table 9.3 ANOVA results for phonemes scores in the CNC word test for
the 5 subjects who participated in Experiment 3. .................................... 177
Table 9.4 ANOVA results for consonant scores in the CNC word test for
the 5 subjects who participated in Experiment 3. .................................... 179
Table 9.5 ANOVA results for fricative scores in the CNC word test for
the 5 subjects who participated in Experiment 3. .................................... 181
Table 9.6 ANOVA results for vowel scores in the CNC word test for
the 5 subjects who participated in Experiment 3. .................................... 183
Table 9.7 ANOVA results for the consonant test for the 5 subjects who
participated in Experiment 3.................................................................... 186
Table 9.8 ANOVA results for the sentence in noise test for the 5 subjects
who participated in Experiment 3. ........................................................... 190
Table 9.9 Summary of word score results in quiet for the current and
previously reported studies. .................................................................... 198
Table 9.10 Summary of speech testing in noise for the current and
previously reported studies. .................................................................... 201
Table 9.11 Confusion matrices for the consonant test in the CI alone
condition.................................................................................................. 203
Table 9.12 Confusion matrices for the consonant test in the CIHA/s
condition.................................................................................................. 204
-17-
Chapter 1
Introduction
Sensorineural hearing loss causes a number of hearing abilities to deteriorate
for the hearing-impaired individual. Typically, the threshold of hearing is
increased much more than the threshold of discomfort. As a result, the level
difference between discomfort and the threshold of audibility, or the dynamic
range of an ear, is much less than that of a normally-hearing ear (Steinberg et
al., 1937). Another difficulty facing listeners is separating sounds of different
frequencies. This is termed frequency resolution or selectivity. This results in
the brain having difficulty in separating a signal from background noise. Intense
sounds can also mask weaker sounds that immediately precede or immediately
follow them (temporal resolution). This happens to a greater extent for hearingimpaired individuals and can adversely affect speech intelligibility.
Each of the above aspects of hearing loss can work in combination resulting in
the hearing-impaired individual understanding much less than a normal-hearing
person. However, the most obvious effect of hearing impairment is a decrease
in audibility due to threshold elevation. Individuals with severe losses may not
hear any speech sounds unless they are at high intensity levels. Unfortunately,
even listeners with normal hearing have poorer speech understanding when
listening at high intensity levels. The hearing-impaired person will have
difficulties with speech understanding not only because of a loss in audibility but
also because of the need to listen at high intensity levels.
Chapter 2 provides a brief background of the normal auditory system and how
this system codes and perceives sound. Any damage to the cochlea within this
system negatively affects the mechanisms required for hearing. As a general
rule, the greater the hearing loss, the larger the amount of amplification needed
to provide the listener with adequate audibility of speech sounds. However, this
rule has certain limitations for individuals with a severe loss in the high
frequencies, but who have good hearing sensitivity in the low frequencies. For
- 18 – Chapter 1 Introduction
these individuals, conventional hearing aids are limited in their ability to provide
adequate high-frequency audibility. This constraint has come about for various
reasons. The hearing-aid fitting must provide a balance between providing
audibility whilst maintaining a signal that is comfortable for the listener.
Secondly, acoustic feedback will often limit the amount of usable gain,
dependent on individual characteristics, such as how effectively the ear canal is
sealed by the earmold. Finally, the typical hearing aid’s upper frequency limit
may not be high enough to ensure the inclusion of certain high-frequency
fricatives in the signal, such as /s/. For example, Stelmachowicz et al. (2001)
evaluated fricative perception for a group of hearing-impaired subjects under
various low-pass filter conditions. To obtain optimum performance for female
and child speakers, it was found that the bandwidth was required to be at least
9000 Hz.
Some hearing-impaired listeners do not obtain speech perception benefits from
high-frequency information (Ching et al., 1998; Hogan et al., 1998; Murray et al.,
1986; Rankovic, 1991). Chapter 3 reviews the role of audibility and speech
perception abilities for sensorineural hearing losses. What remains unclear is
why only certain individuals show this lack of benefit. One explanation is that of
“dead regions”. The term dead region was coined to define a certain region of
the cochlea with non-functioning inner hair cells (Moore et al., 1997; Thornton et
al., 1980). Whereas individuals with no dead regions should obtain a benefit
from high-frequency amplification, individuals with dead regions should show no
benefits from high-frequency amplification. The Threshold-Equalizing Noise
(TEN) test (Moore, 2001; Moore et al., 2000a) was developed as a means for
clinicians to be able to identify dead regions and thereby a means for deciding
when to provide high-frequency amplification. It is based upon the detection of
sinusoids in the presence of a broadband noise. Normally-hearing listeners and
listeners with hearing impairment but without dead regions should show almost
equal masked thresholds (in dB SPL) over a wide frequency range with the test.
Individuals with dead regions should show abnormally high-masked thresholds
at one or more frequencies.
The TEN test was validated in studies where vowel-consonant-vowel (VCV)
nonsense syllables were presented to subjects under various low-pass filter
-19conditions (Baer et al., 2002; Vickers et al., 2001). It was found that consonant
identification generally improved with increasing filter cutoff frequency for
subjects with no dead regions. For subjects with dead regions, performance
improved with increasing cutoff frequency until the cutoff frequency was
somewhat above the estimated edge frequency of the dead region, but hardly
changed with further increases. Unfortunately, these studies all shared one
confounding factor, that of audibility. On average, subjects who showed dead
regions had more severe hearing losses than those subjects without dead
regions. This may have resulted in insufficient audibility of the speech stimuli for
those subjects with dead regions at certain frequencies. This lack of audibility
was likely to have contributed to the poor scores these subjects obtained.
Chapter 3 discusses current literature on cochlear dead regions in further detail.
This thesis describes three experiments which were carried out with hearingimpaired listeners with severe high-frequency hearing loss. Experiment 1 aimed
to investigate dead regions further by reducing the confounding factor of
audibility (Simpson et al., 2005a). The methods used and results found for
Experiment 1 are described in Chapter 7. The presence of dead regions was
evaluated in 10 subjects using the TEN test. All subjects had moderatelysloping high-frequency losses. Results with the TEN test were compared with
subjects’ scores on a consonant test which had various low-pass filter
conditions. Audibility of the consonant stimuli was maximized by carrying out
testing using a direct audio input method. This enabled the level of the speech
stimuli to be increased as required without acoustic feedback occurring.
Regardless of whether it could be shown which individuals showed
improvements from increasing high-frequency audibility, it is clear that current
amplification techniques are not always sufficient at providing high-frequency
information. Alternative sound-processing options, which preserve useful lowfrequency audibility and provide additional high-frequency information, were
investigated further for these individuals. Two such options which were selected
for further investigation in the current study were: frequency-lowering hearing
aids, and cochlear implantation using a modified surgical technique.
A review of sound-processing schemes that have been developed with the aim
of presenting information from high-frequency regions of speech at lower
- 20 – Chapter 1 Introduction
frequencies are described in Chapter 4. Frequency lowering (also known as
“frequency shifting” or “transposition”) was one way of providing this extra
information acoustically. The performance of a novel frequency-compression
device was investigated using tests of speech understanding in quiet (Simpson
et al., 2005c). The device compressed frequencies above a programmable cutoff, resulting in those parts of the input signal being shifted to lower frequencies.
Below the cut-off, signals were amplified without frequency shifting. Seventeen
experienced hearing-aid users with moderate-to-profound sensorineural hearing
loss and sloping audiograms participated in the study. Of the 17 subjects, eight
obtained a significant (p < 0.05) phoneme score increase with the experimental
frequency-compression scheme, eight subjects showed no significant change in
scores, and one subject showed a significant score decrease with the
experimental scheme.
The experimental scheme described above proved successful in providing
additional high-frequency information for some listeners with moderately-sloping
losses whose hearing thresholds became severe-to-profound in the 2000 –
8000 Hz region. However, the scheme was not tested on listeners with steeplysloping hearing losses. Typically, these individuals have near-normal hearing
thresholds in the low frequencies and a profound hearing loss for frequencies
above approximately 1500 Hz. Experiment 2 of the current study examined
whether the experimental frequency-compression scheme described by
Simpson et al. (2005c) would benefit listeners with steeply-sloping hearing
losses (Simpson et al., 2006). The methods used and results found for
Experiment 2 are described in Chapter 8. Seven subjects were included in the
study. All subjects were initially fitted with a conventional device. Speech
perception comparisons in both quiet and noise were made between this device
and the frequency-compression device. The Abbreviated Profile of Hearing Aid
Benefit (APHAB) questionnaire (Cox et al., 1995) was given to subjects at the
end of the trial in order to determine their subjective preferences.
Cochlear implantation is an alternative consideration for individuals with severe
high-frequency hearing loss. Chapter 5 provides a broad outline of how
cochlear implant systems work. Many cochlear implant candidates now have
some measurable unaided hearing thresholds prior to implantation. It is likely
-21that some, if not all, of this residual hearing may be lost as a direct result of
implantation (Boggess et al., 1989; Rizer et al., 1988). The main factors thought
to result in possible loss of hearing include electrode insertion trauma during
surgery, and growth of tissue post-operatively. However, the degree of hearing
loss and the chance for hearing preservation after implantation remain unclear.
Despite evidence that residual hearing may be preserved in some cases (Gantz
et al., 2005; Shin et al., 1997; Skarzynski et al., 2003), partial or complete loss
of residual hearing is a risk clinicians must weigh against the potential
perceptual benefits provided by implantation. For individuals who have a
severe-to-profound hearing impairment across most of the audible frequency
range, a cochlear implant is likely to be beneficial overall, even with the loss of
acoustic sensitivity that usually results from implantation. However, individuals
who have relatively good hearing sensitivity at the lower frequencies are
unlikely to choose cochlear implantation even if they have found conventional
hearing aids to be of little benefit. At the present time, they are not considered
candidates for conventional cochlear implantation since their residual lowfrequency hearing and relatively good speech perception abilities, compared to
conventional cochlear implant candidates, precludes them.
Ongoing studies (Gantz et al., 2003; Gantz et al., 2004; Gantz et al., 2005;
Turner et al., 2004) are exploring cochlear implantation with this group of
hearing-impaired listeners. Essentially these studies have two broad aims. One
aim is finding a surgical method which preserves low-frequency residual hearing
when inserting an electrode array. If this low-frequency hearing can be
preserved, the second goal is to determine the optimum means of combining
the acoustic sound from the listeners’ natural hearing together with the electric
stimuli from the implant speech processor.
With these goals in mind, ongoing collaboration at the University of Iowa and
other international research centres with Cochlear Ltd. has resulted in the
development of a short-electrode array. The electrode array is shorter than a
conventional electrode array (10 mm versus 24 mm), has a smaller diameter,
and has 6 active electrodes. Of the 24 subjects who participated in the study, 22
maintained some useful acoustic hearing in the implanted ear post-operatively
(Gantz et al., 2005). At 6 months post-implantation, monosyllabic word
- 22 – Chapter 1 Introduction
understanding when compared with the pre-operative binaural hearing-aid
condition was significantly improved for 10 of the 11 subjects tested (Gantz et
al., 2005).
Other reported studies have used a standard electrode array but varied the
insertion depth during surgery (Gstoettner et al., 2004; Kiefer et al., 2005). Of
the 34 subjects who participated in these studies, a total of 29 subjects
maintained some useful residual hearing in the implanted ear post-operatively.
Speech perception results were reported for 14 subjects. All of these subjects
showed improvements in speech scores post-operatively. Chapter 6 describes
implantation outcomes for patients with residual hearing in further detail.
Experiment 3 of the current study investigated the fitting of a cochlear implant
together with aided residual hearing by means of matching frequency and/or
perceived pitch between acoustic and electric modalities. The methods used
and results found for Experiment 3 are described in Chapter 9. Five subjects
with steeply-sloping losses were implanted with a standard electrode array. A
modified surgical technique was used in an attempt to preserve residual
hearing. If useful hearing was present post-operatively, subjects were fitted with
an ITE hearing aid together with the speech processor in the implanted ear. All
subjects were asked to continue wearing a hearing aid in the contralateral ear.
The characteristic frequency in the cochlea corresponding to the most apical
electrode was determined with psychophysical measures. This frequency
information was used to create individual fitting programs for each of the
subjects which attempted to mimic the normal ear’s frequency-place
relationship. Each subject tried two different fitting programs. The first program
assigned frequency channels to the electrode array which attempted to mimic
the cochlea’s frequency-place relationship. After 12 weeks, each subject’s
implant fitting program was changed. For the second program electrodes were
assigned to the full-frequency range. Subjects wore this map for a further 12
weeks. Speech testing in both quiet and noise was carried out for both
programs. Chapter 10 provides a general discussion about all three
experiments as well as final conclusions.
-231.1
Aims and Objectives
The main aims and objectives of the current study can thus be summarized as
follows:
1. To evaluate whether high-frequency amplification benefits listeners with
severe-sloping hearing losses.
2. To evaluate whether useful high-frequency information can be provided to
listeners with severe high-frequency hearing loss by either a frequencycompression hearing aid or a cochlear implant.
3. To evaluate the effect of providing electric hearing in regions of profound
high-frequency sensorineural hearing loss together with acoustic hearing in
regions of normal-to-moderately impaired low-frequency hearing in individuals
with suitable types of hearing impairment.
4. To optimize the fitting of a cochlear implant together with aided residual
hearing, in particular by means of matching frequency and/or perceived pitch
between acoustic and electric modalities.
- 24 – Chapter 1 Introduction
- 25 -
Chapter 2
Cochlear mechanisms and processes
2.1
Introduction
Cochlear hearing loss occurs when the structures inside the cochlea are
damaged. This can occur in several different ways, such as exposure to
ototoxic chemicals or intense sounds, infection, metabolic disturbances,
allergies, or as a result of genetic factors. The damage may extend beyond the
cochlea to include higher centres in the auditory pathway. The more general
term of sensorineural hearing loss is used to describe a hearing loss where
damage occurs to either the cochlea or neural structures.
This chapter aims to provide a brief background of how the auditory system
codes and perceives sound, in particular the role of the basilar membrane and
cochlear hair cells. The concepts of cochlear tuning and frequency selectivity
are outlined, which will provide the basis for the following chapters.
2.2
Anatomy of the inner ear
The human auditory system comprises the ears and their connections to and
within the central nervous system. It can be divided into the outer, middle, and
inner ears, the auditory nerve, and the central auditory pathways. Figure 2.1
shows the major anatomical structures of the ear.
The inner ear begins at the oval window. It consists of sensory organs of
hearing (the cochlea) and of balance (the semicircular canals, utricle, and
saccule). The balance system is also known as the vestibular system. The
vestibule, cochlea and vestibular apparatus are located beyond the oval
window. The cochlea forms a cone-shaped spiral with about 2 and ¾ turns. It is
widest at the base and tapers towards the apex. It is divided along its length by
two membranes, Reissner’s membrane and the basilar membrane, which divide
the cochlea into three chambers: the scala media, scala vestibuli, and scala
tympani. The Organ of Corti rests on the basilar membrane in the scala media.
The structures and orientation of the Organ of Corti are shown schematically in
- 26 – Chapter 2 Cochlear mechanisms and processes
Figure 2.2. It consists of a single row of inner hair cells (IHCs), three to five rows
of outer hair cells (OHCs), various supporting cells, and the pillar cells forming
the tunnel of Corti. This tunnel separates the OHCs from the IHCs.
Figure 2.1 Cross- sectional view of the human ear. The ear is divided into an outer,
middle and inner section. The outer ear consists of the pinna and external auditory canal.
The middle ear consists of the tympanic membrane, an air-filled space, and three
ossicles (malleus, incus and stapes). The inner ear consists of the cochlea and the
organs of balance. Figure from http://www.uic.edu/classes/psych/psych352jw/c5.html.
Figure 2.2 Cross section of the Organ of Corti as it appears in the basal turn of the
cochlea. Figure from http://ourworld.compuserve.com/homepages/dp5/evod3.htm.
- 27 There are approximately 3500 IHCs. These are supported by phalangeal cells
which hold the rounded base of the IHC. The outer hair cells are shaped like
test tubes and are supported by Dieter Cells. There are about 12 000 OHCs in
the cochlea (Gelfand, 1998). The structures and organization of the inner and
outer hair cells have been described in great detail (Lim, 1980; Lim, 1986). They
differ from each other structurally, as well as functionally. The flask-shaped
IHCs contain an extensive system of tubulovessicular endoplasmic reticulum,
Golgi apparatus and mitochondria. This suggests a high level of metabolic
activity, which may be related to the transduction of mechanical to
electrochemical energy. Their endoplasmic reticulum is far less developed than
in the OHCs, which contain a highly organized endoplasmic reticulum system.
The OHCs also contain contractile proteins in their cell bodies and cell
membranes, stereocilia and cuticular plates (Slepecky et al., 1988) which
enable each OHC to actively contract and elongate.
The upper surface of both inner and outer hair cells contains a thickening called
the cuticular plate which is topped by stereocilia, and a noncuticular area that
contains the basal body of a rudimentary kinocilium. Each OHC contains as
many as 150 stereocilia. These stereocilia increase in number from about 58 at
the cochlear apex to 150 at the base (Wright, 1981). The OHC stereocilia are in
contact with the overlying tectorial membrane. Each inner hair cell has roughly
50-70 stereocilia. These are thicker than those on the OHCs (Lim, 1980). Their
numbers do not appear to change along the length of the cochlea. Also, the IHC
stereocilia appear to have no or minimal contact with the tectorial membrane.
2.3
Perception of sound in the auditory system
Sound waves travel along the auditory canal and set the tympanic membrane
into vibration. These vibrations are transmitted through the middle ear and its
three ossicles, the malleus, incus and stapes, to the oval window at the base of
the cochlea. Figure 2.3 shows the way acoustic signals are passed from the
basilar membrane to the auditory nerve. Briefly, a vibration is produced along
the basilar membrane by the vibrations delivered to the oval window. The
displacement of the basilar membrane results in a shearing force and the
tectorial membrane moves sideways. As a result, the stereocilia on top of the
hair cells also move sideways.
- 28 – Chapter 2 Cochlear mechanisms and processes
auditory nerve
spiral
ganglion cell
release of
neurotransmitter
deflected
stereocilia
incoming
sound
hair cell
basilar membrane
high frequency
low frequency
Figure 2.3 Representation of the way acoustic signals are passed from the basilar
membrane to the auditory nerve. Sound waves at a certain frequency cause the basilar
membrane to be displaced at a particular place. The basilar membrane responds best or
maximally to this characteristic frequency. Information about this signal is conveyed via
the haircells and the spiral ganglion cells to the auditory nerve.
- 29 This sideways motion is via direct contact with the tectorial membrane for the
OHCs, and via the fluid drag in the Organ of Corti for the IHCs. Bending of the
IHC stereocilia causes an excitatory response. Transduction pores on the
stereocilia open allowing access to a channel across which ions can flow.
Mechanical energy is thus converted into electrical energy. A chemical
mediator, or neurotransmitter, is transmitted across the space between the hair
cell and the afferent nerve. The neurotransmitter causes an excitatory
response, or action potential along the auditory nerve. The auditory nerve then
carries the signal through the central auditory system to the auditory cortex in
the brain. Thus, the IHCs play an important role in transducing mechanical
movements in the cochlea into neural activity. They are responsible for
conveying information about sound from the cochlea onto higher auditory
pathways via afferent neuron connections. The OHCs are part of an “active
mechanism” in the cochlea. By changing their length, shape and stiffness in
response to sound they can in turn influence the response of the basilar
membrane. The following section describes some of the processes that occur at
the basilar membrane in more detail.
2.3.1 Cochlear tuning and frequency selectivity
The following section describes the response of the basilar membrane to
stimulation from a pure tone. When the tone sets the oval window into motion, a
pressure difference is created between the upper and lower surface of the
basilar membrane. The wave travels rapidly through the fluids of the cochlea.
As a result, the pressure difference is applied almost simultaneously along the
length of the basilar membrane. The travelling wave moves along the basilar
membrane from its base to apex. The amplitude of the wave increases at first
and then decreases rather quickly. The mechanical properties of the basilar
membrane affect how it responds to sounds of different frequencies. The basilar
membrane is organized in terms of place of displacement and frequency. This
tonotopic organization results in displacement of the membrane reaching a
peak near the apex of the cochlea in response to low frequencies and near the
base for higher frequencies. This is known as tuning. A certain point along the
basilar membrane responds best, or with maximum displacement, to a
- 30 – Chapter 2 Cochlear mechanisms and processes
particular frequency. This frequency is known as the characteristic frequency
(CF).
Auditory nerve fibres also have highly-tuned acoustic frequency response
characteristics, or frequency selectivity, as they respond maximally to certain
frequencies. Each nerve fibre has a CF, which is the frequency at which its
threshold is lowest (or at which the greatest firing rate is produced). The CF
corresponds closely to the frequency which causes maximum displacement of
the basilar membrane at the point where the hair cell is located, which the
neuron fibre innervates.
The tuning of the auditory fibres is shown in Figure 2.4. Each fibre will respond
to a range of frequencies if the stimulus level is high enough. This frequency
range extends considerably below the CF, but is restricted above it. A fibre will
thus respond to intense stimulation at frequencies below its CF, but is only
minimally responsive to stimulation above it. The fibres are more narrowly tuned
to a particular frequency at lower intensity levels (Kiang, 1965). The frequency
of a sound can therefore be coded in the auditory system in terms of the place
in the cochlea of maximum neural excitation.
The pattern of basilar membrane displacement is highly sensitive to the health
of the cochlea. Studies have shown that the mechanical response of the basilar
membrane consists of two components: a) a low-pass filter type of response,
and b) a sharply-tuned response (Khanna et al., 1982; Khanna et al., 1986a;
Khanna et al., 1986b; Sellick et al., 1982). This is shown in Figure 2.5 where the
solid black line represents the shape of a normal auditory neural tuning curve. It
was found (Khanna et al., 1986a; Khanna et al., 1986b) that the low-pass filter
response appeared to be relatively immune to trauma whereas the second
component was highly susceptible to trauma. Although both inner and outer hair
cells were needed to produce a normal tuning curve, the characteristics of the
sharply-tuned response appeared to be dependent on the condition of the
OHCs.
The implication of these findings was that the tip component of the cochlea’s
mechanical tuning curve was associated with the OHCs and in particular the
condition of their stereocilia (Liberman, 1984; Liberman et al., 1984a; Liberman
et al., 1984b; Liberman et al., 1984c; Liberman et al., 1984d). Another important
- 31 observation was that of Liberman (1984) who studied how cochlear damage
due to acoustic trauma affected the threshold and characteristic frequency (CF)
of the neural tuning curve. The CF was shifted downwards as threshold
sensitivity worsened, at least for neurons with an original CF above 1000 Hz.
The CF of the mechanical tuning curve became lower in frequency in cochleas
having greater amounts of OHC damage.
Number of spikes per second
Response to low intensity sound
Response to high intensity sound
Frequency (Hz)
Figure 2.4 Schematic representation of the tuning of an auditory nerve fiber. The
frequency (x-axis) at which the greatest firing rate (y-axis) is produced is known as the
characteristic frequency. It is represented by the dashed line. High intensity sounds
(dotted line) result in the nerve fiber responding to a larger frequency range, whereas
low-intensity sounds (solid line) result in the nerve fiber responding to a narrower
frequency range.
According to Liberman et al. (1986), significant damage to the IHC stereocilia is
almost always associated with reduced sensitivity on the tail of the tuning curve.
Liberman et al. (1986) found no cases of IHC damage without the OHCs also
being affected. It appears that the OHCs are more vulnerable to insult than the
IHCs, and that damage to the IHCs is nearly always associated with damage to
the OHCs. However, some evidence of IHC lesions have been found, even in
the presence of normal OHCs (Borg et al., 1995; Engstrom, 1983). Damage
may depend on the type of noise the listener is exposed to. High-level, short-
- 32 – Chapter 2 Cochlear mechanisms and processes
term noise results mainly in IHC stereocilia damage, whereas low-level, longterm noise seems to affect the OHCs.
Outer hair cell damage
Sound pressure level
Inner hair cell damage
Frequency
Figure 2.5 Schematic representation of an auditory-nerve tuning curve. The solid black
line shows the normal condition. The dotted line shown by the illustration on the left
show how inner hair cell damage results in an overall loss of sensitivity. The dotted line
shown by the illustration on the right shows the change in the tuning curve as a result of
outer hair cell damage.
- 33 -
In summary, it seems likely that, for many types of cochlear
hearing loss, the primary cause is damage to the OHCs and/or
IHCs. The stereocilia may be distorted or destroyed, or entire
cells may die. The OHCs are usually, although not necessarily,
affected first. Sensitivity to weak sounds is reduced, the tuning
curves on the basilar membrane become more broadly tuned,
and the frequency-selective effects weaken or disappear
altogether. We can assume that mild hearing losses are
primarily the result of damage to the OHCs. For more severe
losses, it is likely that both the OHCs and IHCs are damaged. It
is a possibility that the IHCs at certain places along the basilar
membrane may be completely non-functioning. In such cases,
vibration at those places is not detected by the auditory fibres
normally innervating that place.
The term “dead region” was coined by Moore & Glasberg
(1997) to refer to an area of the basilar membrane over which
there are no functioning IHCs.
Dead regions will be discussed further in Chapter 3.
- 34 – Chapter 2 Cochlear mechanisms and processes
- 35 -
Chapter 3
The role of audibility in speech perception
3.1
Introduction
Damage to the cochlea generally results in a number of hearing abilities
deteriorating in individuals with a sensorineural hearing impairment. However,
the most obvious result of hearing impairment is a decrease in audibility due to
threshold elevation. Individuals with severe losses may not hear any speech
sounds unless they are presented at high intensity levels.
This chapter reviews how audibility affects speech perception abilities for
listeners with sensorineural hearing losses. Past research (summarized in
section 3.3 of the current chapter) differs in the methods used for defining the
relationship between speech perception scores and audibility of the signal.
Some studies have compared alternative amplification strategies where the
strategies differ in the amount of amplification given to the listener. Others have
made use of the Speech Intelligibility Index or SII. The SII provides a means of
calculating the relationship between audibility and speech perception. The SII
will be described in section 3.2 of the current chapter. It will be shown that, for
some listeners, increased audibility of high-frequency information resulted in a
decrease in speech perception. One of the possible reasons for this lack of
benefit of high-frequency amplification, when found, may be the presence of
non-functioning IHCs over a certain region of the cochlea (Moore et al., 1997).
Such a region is referred to as a “dead region”. The concept of dead regions,
their diagnosis and implications will be discussed in sections 3.4 and 3.5 in the
chapter below.
3.2
The Speech Intelligibility Index (SII)
The relationship between audibility and speech recognition can be quantified
using the Articulation Index (French et al., 1947), now called the Speech
Intelligibility Index (SII). Numerous studies have used the SII as a tool to predict
speech perception scores (Dubno et al., 1989a; Dubno et al., 1989b; Humes et
al., 1992; Humes et al., 1986; Kamm et al., 1985; Pavlovic et al., 1984; Zurek et
- 36 – Chapter 3 The role of audibility in speech perception
al., 1987). The ANSI S3.5 (ANSI, 1969; ANSI, 1997) SII calculation, based on
the original French & Steinberg (1947) method, has been used in most
published SII studies.
It is defined by the equation:
n
SII = P ∑ I i × Wi
i =1
Equation 1
where n is the number of frequency bands, Ii and Wi are the values of the
importance function (I) and the audibility function (W) associated with the
frequency band i, and P is the proficiency factor. The relative contribution of
different frequency bands to speech perception is represented by the frequency
importance function (I). This function’s shape will change depending on the
composition of the speech material, the recording and presentation of the
speech material, and the experimental methods and data evaluation
procedures. The audibility function (W) has a value between 0.0 and 1.0 and
characterizes the effective proportion of the speech dynamic range which is
audible within each band. The ratio between the SII predicted from
measurements of the speech spectrum and the SII derived from a subject’s
speech recognition score is the proficiency factor (P). The calculated SII, a
value between 0.0 and 1.0, may be interpreted as the proportion of total
information present in the speech signal that is effectively processed by the
listener. Speech perception scores and the SII are related by a monotonically
increasing function known as a transfer function. The transfer function is
dependent on the speech material selected as well as the speaker. This results
in the researcher having to derive a unique transfer function for any speech test
which is used in SII calculations.
3.3
Audibility studies with hearing-impaired listeners
Amplification aims to provide an audible signal over a wide frequency range for
a hearing-impaired listener. As a general rule, this would involve providing the
most gain at frequencies which have the greatest hearing loss. In the case of
individuals with high-frequency hearing loss, the largest amount of amplification
- 37 would occur in this frequency range. However, research indicates that some
individuals with moderate-to-severe hearing impairment do not benefit from
amplification at high frequencies (Hogan et al., 1998; Murray et al., 1986;
Rankovic, 1991).
Murray & Byrne (1986) fitted hearing aids to 5 subjects with sloping highfrequency hearing losses. The amplification was shaped to suit their hearing
losses but had variations in the high-frequency cut-off selected (1500, 2500,
3500, and 4500 Hz). Three of the subjects obtained the highest speech
recognition scores with bandwidths below the widest bandwidth available. Of
the two remaining subjects, one showed a small improvement in speech scores
with the highest bandwidth of 4500 Hz when compared to the 3500 Hz
condition. The fifth subject’s scores were highly variable and inconclusive.
A similar study by Rankovic (1991) compared two different amplification
strategies to a procedure that attempted to maximize audibility of speech
(AImax). Due to limitations of the equipment, maximum audibility was not
achieved for any subject, although the AImax strategy did result in higher
audibility than the other two strategies. All of the ears tested had relatively
normal low-frequency hearing thresholds and sloping high-frequency hearing
losses. Of the 13 ears tested, there were no cases where AImax significantly
improved performance over both alternative strategies. The majority showed no
decrease in performance with the AImax condition. Four ears performed more
poorly with AImax when compared to conditions where audibility was less than
the maximum possible. Even though AImax generally provided the highest gain,
it did not provide the highest speech recognition score. These studies seemed
to indicate that the widest bandwidth, or greatest audibility, is not always
desirable in an amplification system.
Four normal-hearing and five hearing-impaired subjects participated in a trial
conducted by Turner & Robb (1987). Consonant-vowel (CV) tokens of the 6
stop consonants /b,d,g,p,t,k/ were presented with the vowel /a/ in the
soundfield. Recognition scores were obtained across a wide range of
presentation levels with a maximum level of 100 dB SPL. As expected,
consonant recognition performance improved as audibility increased for normalhearing subjects. In contrast, the results of the hearing-impaired group showed
- 38 – Chapter 3 The role of audibility in speech perception
that in most cases equivalent levels of audibility for the two groups did not result
in equivalent levels of recognition. This was especially true for those subjects
with high-frequency hearing loss. However, stimuli were not fully audible for the
hearing-impaired subjects, even at the highest presentation level. It is possible
that further increases in audibility would have resulted in improved recognition
as many of the hearing-impaired subject group’s scores were still rising at the
maximum level.
These results are supported by Hogan & Turner (1998) who investigated the
benefits of providing high-frequency information to hearing-impaired listeners.
Nonsense syllables were low-pass filtered at various cut-off frequencies and
presented to 5 normal-hearing and 9 hearing-impaired listeners at various
presentation levels. The AI was used to quantify audibility for each condition
and for each listener. Listeners with mild high-frequency loss performed
similarly to the normal-hearing group with scores improving as audibility
increased. Recognition scores of those subjects with severe losses, however,
showed that increasing the audibility resulted in no further improvements in
scores. The hearing-impaired listeners were not making use of high-frequency
information above 4000 Hz to improve their speech recognition, in some cases
where the hearing loss was less than 55 dB HL. Again, it is possible that, even
at the highest presentation level, some components of the speech signals
remained inaudible to some subjects.
Dubno et al. (1989b) made use of articulation index (AI) theory to evaluate stopconsonant recognition of 18 normal-hearing and 10 hearing-impaired listeners.
The actual and predicted recognition scores were compared. The procedures to
calculate the AI by French & Steinberg (1947) were followed with some
modifications. A unique transfer function and frequency importance function
were determined for the 9 stop consonant–vowel syllables used as test stimuli.
It was found that the AI tended to overestimate performance for certain hearingimpaired subjects. The accuracy of the prediction seemed to decrease as the
hearing threshold increased for subjects with high-frequency hearing loss. The
AI prediction closely estimated performance for mild losses but overestimated
performance for moderate-to-severe sloping losses. Dubno et al. (1989b)
hypothesized that the AI calculation was not sensitive enough to changes in
- 39 hearing threshold sensitivity if hearing threshold changes occurred steeply over
a narrow frequency range.
Ching et al. (1998) examined the relationship between audibility and speech
recognition scores for a group of 54 subjects. Fourteen of the subjects had
normal hearing, 19 had mild-to-moderate losses, and 21 had severe-toprofound losses. Speech perception performance on sentences and nonsense
syllables at various sensation levels were compared with SII predictions. For the
majority of hearing-impaired listeners, the SII underestimated speech scores
obtained at low-sensation levels. For high-sensation levels, the SII prediction
was close to accurate for listeners with mild-to-moderate losses, but greatly
overestimated speech scores for listeners with severe-to-profound losses. The
authors carried out further investigation by analysing the data with individual
frequency-dependent proficiency factors. Specific frequencies were expressed
in terms of degree of contribution of the audible signal to intelligibility. For
listeners with severe losses, the proficiency factor often became zero or
negative at high frequencies. Specifically, an audible signal in the 2800 – 5600
Hz region made no contribution to intelligibility for severe losses greater than 80
dB HL. This together with other studies (Hogan et al., 1998; Murray et al., 1986;
Rankovic, 1991) raises the question of whether amplification should be provided
in the high frequencies if the loss is severe.
All of the studies described above have been carried out for testing in quiet. It is
of interest to determine whether similar outcomes have been found for testing in
noise. In Turner & Henry (2002) speech together with multi-talker babble was
presented to listeners with varying degrees of hearing loss. The signal was lowpass filtered and speech recognition was measured as additional highfrequency speech information was provided to the hearing-impaired listeners. It
was found that for all subjects, regardless of hearing loss or frequency range
that an increase in cut-off frequency resulted in an increase in recognition
score. However, the amount of improvement in recognition was small, and the
maximum score obtained by most listeners was low. In a similar study, 10
normal-hearing subjects were compared with 10 subjects with sloping
sensorineural hearing losses (Hornsby et al., 2003). Sentence recognition in
noise was presented to both groups at various cut-off frequencies. Small
- 40 – Chapter 3 The role of audibility in speech perception
improvements in intelligibility were found for the hearing-impaired group. For
this group of subjects, the SII generally over-predicted performance for
frequencies higher than 1200 Hz. These studies in noise suggested that
hearing-impaired listeners were able to use amplified speech information in
background noise, but the magnitude of improvement was small.
In summary, the SII appears to provide good prediction of speech recognition
scores for normal-hearing listeners and those with mild and moderate hearing
loss (Dubno et al., 1989a; Dubno et al., 1989b; Humes et al., 1986; Kamm et
al., 1985; Pavlovic et al., 1984; Zurek et al., 1987). However, the SII tends to
overestimate the performance of those hearing-impaired subjects with severe
high-frequency loss (Ching et al., 1998; Dubno et al., 1989b; Kamm et al., 1985;
Pavlovic, 1984). For these listeners, the SII method predicts a better score for a
given degree of audibility than the hearing-impaired listener actually obtains.
Research indicates that a small number of individuals with moderate-to-severe
hearing impairment do not benefit from amplification at high frequencies (Ching
et al., 1998; Hogan et al., 1998; Murray et al., 1986; Rankovic, 1991; Turner et
al., 1987), at least for quiet conditions.
It has been suggested that this lack of benefit of high-frequency amplification,
when found, is due to the presence of non-functioning IHCs over a certain
region of the cochlea (Moore et al., 1997). Such a region is referred to as a
“dead region”.
3.4
The definition of a dead region
Chapter 2 (see section 2.3.1) discussed how each nerve fibre has a
characteristic frequency (CF). The CF corresponds closely to the frequency
which causes maximum displacement of the basilar membrane at the point
where the hair cell is located which the neuron fibre innervates. However, if the
stimulus level is high enough, each fibre will respond to a range of frequencies.
Consider now a case of cochlear hearing loss which is associated with damage
to the IHCs. Basilar membrane vibration in the area of damaged hair cells will
not be detected via the neurons directly innervating that region. However,
intense signals can be detected via adjacent IHCs and neurons. For this
reason, Moore (2001) defines a dead region in terms of the CFs of IHCs and/or
neurons immediately adjacent to the dead region. If the dead region extended
- 41 to the IHCs corresponding to higher frequencies, then a high-frequency sound
would be detected via neurons that are tuned to lower frequencies via a
downward spread of excitation. This detection of a tone of a particular
frequency via IHCs and neurons with CFs different from that of the tone is
referred to as “off-frequency listening”.
3.4.1
Effects of off-frequency listening
One result of off-frequency listening is that an audiogram may give a misleading
impression. If a listener has an area of non-functioning IHCs, a true threshold
should be measurable on the audiogram in the frequency range corresponding
to the area of damage in the inner ear. But due to the distribution of the nerve
fibres, if a tone in the frequency range corresponding to the area of damage
was intense enough, adjacent CFs would respond. Thus, the audiogram may
indicate a measurable hearing threshold when actually the “true” hearing
threshold in a dead region is unmeasurable.
Absolute thresholds shown on an audiogram can provide some clues as to
whether a dead region may be present. Chapter 2 discussed how severe
hearing loss is likely to consist of both inner and outer hair cell damage. It has
been shown that OHC damage results in reduced basilar membrane vibration
for sounds with low levels (Ruggero, 1992). A sound then has to be presented
at higher than normal intensity levels in order to reach equivalent amounts of
basilar membrane vibration. It is of interest to determine what effect OHC
damage would have on threshold measurement. This has been described in
terms of “gain”. Gain represents “the difference in input sound level needed to
produce a fixed low amount of vibration on the basilar membrane in a healthy
ear and an ear in which the active mechanism is not functioning” (Moore, 2001).
Physiological data obtained from animals suggests that the maximum gain
provided by the OHCs is roughly 65 dB at high frequencies (Ruggero et al.,
1997). This implies that absolute thresholds up to 65 dB higher than normal
could be caused by either “pure” OHC damage or a combination of OHC and
IHC damage. Absolute hearing thresholds greater than 65 dB in the high
frequencies are probably associated with both IHC and OHC damage. Absolute
thresholds of 90 dB higher than normal are likely to be associated with a dead
region i.e. complete destruction of the IHCs (Moore, 2001).
- 42 – Chapter 3 The role of audibility in speech perception
We can assume that an individual presenting with a severe or profound highfrequency hearing loss is likely to have dead regions, but it is unclear from the
audiogram at which frequency the dead region would begin or end. Diagnosis
and the methods for detecting dead regions are discussed below.
3.5
Methods of detecting dead regions
3.5.1
Psychophysical tuning curves
Psychophysical Tuning Curves (PTCs) are measured by fixing a test signal in
frequency and level, usually just above threshold level. A masking noise is
created using sinusoid or narrow band noise. Figure 3.1 illustrates an example
of PTC measurement.
The level of the masking noise required to just mask the signal is determined for
several masker centre frequencies. For a subject with normal hearing, the
frequency at which the masker level is lowest, or the tip of the PTC, would lie
closest to the signal frequency (Moore, 1978). However, PTC measurement
with hearing-impaired listeners have shown tips shifted away from the signal
frequency (Moore et al., 2001; Thornton et al., 1980; Turner et al., 1983). The
frequency at the tip of the PTC indicates the boundary of the dead region when
the tip is shifted markedly from the signal frequency. If the listener had a highfrequency dead region, the signal would be detected via a downward spread of
excitation. The tip of the PTC would also be shifted downwards towards lower
frequencies.
The consistency between PTC measurements and the TEN test was
investigated in 18 ears with sloping high-frequency hearing loss. It was found
that TEN and PTC results agreed in 56% of cases (Summers et al., 2003). In
eight ears, the two tests provided conflicting results. For these cases, the TEN
test identified dead regions, whereas PTC measurements did not. Results
suggested that the TEN test may not be a sensitive or reliable measure of dead
regions. The findings are discussed in Moore (2004). It is argued that Summers
et al. (2003) PTC results may have been affected by beat detection and
combination tones which may have lead to PTCs with tips at the signal
frequency even when this signal frequency falls within a dead region.
Masker level (dB SPL)
- 43 -
Signal frequency
Shifted PTC tip
Non-shifted PTC tip
Masker frequency (kHz)
Figure 3.1 Schematic representation of PTC results. The dashed line indicates the signal
frequency. A PTC for a normal ear is shown by the dotted line. The tip of the curve (filled
square) is at a frequency close to that of the signal frequency. A PTC for an ear with a
high-frequency dead region is shown by the solid line. The tip of the PTC (filled circle)
has been shifted downwards towards lower frequencies.
PTC measurement does have limitations. The frequency at the tip of the PTC
may not give an accurate estimate of the boundary when the shift is small.
PTCs are also time-consuming to measure and have been used mainly in
laboratory rather than clinical settings. For these reasons the ThresholdEqualizing Noise (TEN) test was developed as a clinical means for diagnosing
dead regions. The test is described fully below.
3.5.2
Masking with Threshold-Equalizing Noise (TEN)
The TEN test (Moore, 2001; Moore et al., 2000a) was developed as a means
for clinicians to be able to detect the presence of dead regions. It is based upon
the detection of sinusoids in the presence of a broadband noise. The noise is
spectrally shaped to produce almost equal masked thresholds (in dB SPL) over
a wide frequency range for normally-hearing listeners and for listeners with
hearing impairment but without dead regions. Individuals with dead regions
show abnormally high-masked thresholds at one or more frequencies. Further
- 44 – Chapter 3 The role of audibility in speech perception
detail about the TEN test is provided in Chapter 7 (see section 7.3.3
Measurement of dead regions).
The TEN test too has some potential limitations. Some hearing-impaired
subjects have thresholds so high that they cannot be measured with a standard
audiometer. It would not be possible to produce masking at these frequencies.
In other cases thresholds are high but still measurable, but the TEN may not be
sufficiently intense to mask all frequencies of interest. The TEN may also prove
to be uncomfortably loud for subjects with sloping-hearing losses. The TEN
would be increased to mask certain frequencies, where the hearing loss is
severe, resulting in discomfort at other frequencies, where the hearing loss is
less severe.
Moore et al. (2000a) assessed the validity of the test by measuring PTCs on 14
hearing-impaired listeners and then performing the TEN test on the same group
of subjects. A good correspondence was found between the results obtained
using the TEN test and the measured PTCs. In other words, abnormally highmasked thresholds in the TEN test were associated with PTCs with shifted tips.
Further studies on the TEN test were carried out. Vowel-consonant-vowel
(VCV) nonsense syllables were presented to 10 subjects under various lowpass filter conditions (Vickers et al., 2001). The limits of any dead region for
each subject were defined by both PTC measurement and the TEN test results.
It was found that consonant identification generally improved with increasing
cutoff frequency for subjects with no dead regions. For subjects with dead
regions, performance improved with increasing cutoff frequency until the cutoff
frequency was somewhat above the estimated edge frequency of the dead
region, but hardly changed with further increases. A few subjects had worsening
performance with further increases in the cutoff frequency, although this
worsening was significant for only one subject.
The percentage information transmission for the phonetic features of voicing,
place, and manner as a function of cutoff frequency was determined for each
subject. The sequential information analysis (SINFA) method was used in which
each feature was treated independently. The percentage of information
transmitted was nearly always highest for voicing and lowest for place. High
scores were found for voicing for subjects with and without dead regions, even
- 45 at very low cutoff frequencies, indicating that information about voicing can be
extracted effectively from low-frequency components of speech. Scores for
manner and place tended to increase with cutoff frequency for those subjects
without dead regions. For subjects with dead regions, scores for manner and
place increased initially with increasing cutoff, but did not improve further once
the cutoff was more than one octave above the estimated edge frequency of the
dead region. The results seem to indicate that there is little benefit in amplifying
frequencies well inside a dead region, but there may be some benefit in
amplifying frequencies up to one octave above the estimated edge frequency of
the dead region. However, a confounding factor in the study was that subjects
with dead regions had greater high-frequency hearing losses than the subjects
without dead regions. This would have resulted in insufficient audibility of the
speech stimuli for those subjects with dead regions at certain frequencies. This
lack of audibility was likely to have contributed to the poor performances these
subjects obtained.
Rankovic (2002) calculated Articulation Index (AI) predictions using the Fletcher
method of calculation (Fletcher, 1953; Fletcher et al., 1950) for the data
published by Vickers et al. (2001). The AI using the Fletcher method is a
product of four factors. The gain factor describes how the AI increases with the
level of the signal. The desensitization factor describes that when the speech
level exceeds a certain value, further increases in level will not result in further
increases in performance. The maximum articulation factor equals the AI when
listening conditions are optimal. And the distortion factor accounts for various
kinds of nonlinear distortion, such as reverberation.
It was found that the AI was generally accurate in predicting the consonant
recognition test scores of subjects irrespective of the presence/absence of dead
regions. For subjects without dead regions, both the AI and consonant
recognition score increased with increasing cutoff frequency. For subjects with
dead regions, there was little or no AI change with increasing low-pass filter
frequency, particularly at cutoffs within the dead regions. This finding implies
that audibility was an important factor in the Vickers et al. (2001) study. The
TEN test did not seem to provide the clinician with any additional information, as
AI calculations alone were able to predict whether a listener would benefit from
- 46 – Chapter 3 The role of audibility in speech perception
high-frequency amplification, at least for this group of subjects. However, the AI
was not accurate in predicting the incremental benefit of amplifying frequencies
well above the estimated edge frequency of a dead region (Moore, 2002). The
AI predicted that speech scores should improve by 10-15%, whereas the largest
improvement found in the Vickers at al. (2001) study was only 7%.
Testing on the same subjects was also carried out using nonsense-syllable
stimuli in noise (Baer et al., 2002). Results were similar to those obtained by
Vickers et al. (2001) with performance improving for cutoff frequencies up to
1.5-2 times the edge frequency of the dead region, but hardly changing with
further increases. To address the issue of audibility, a modification of the AI
(Moore et al., 1998) was used to calculate the audibility of the speech stimuli.
Baer et al. (2002) concluded that frequency components of the stimuli within the
dead regions were at least partially audible to subjects. The authors maintained
that subjects with dead regions do not make as effective use of audible highfrequency speech information as subjects without dead regions.
Further results regarding the TEN test were found by Vestergaard (2003) and
Mackersie et al. (2004). Vestergaard (2003) performed testing on 22 subjects
with moderate-to-profound sensorineural hearing losses. Eleven of these
subjects showed abnormally high-masked thresholds in the TEN test, indicating
a dead region. Estimates of audibility were used to assess the connection
between dead regions and the ability to recognize low-pass filtered speech
stimuli. There was an inconsistent relationship between the benefit of low-pass
filtered speech and the location of the dead region for certain subjects.
Surprisingly, the dead-region subjects showed superior speech recognition
abilities under conditions of poor audibility, compared to no-dead-region
subjects. However, large variability was found in the dead-region group.
Vestergaard (2003) concluded that PTCs may be more appropriate than the
TEN test for accurately estimating the boundaries of a dead region.
In the Mackersie et al. (2004) study, a total of 16 ears with steeply-sloping
hearing losses were tested. Eight of those ears had suspected dead regions
based on the TEN test. The remaining 8 ears were selected as a control group.
This group had similar audiometric thresholds to the test group but had normal
TEN test results. Testing was carried out in the soundfield with subjects wearing
- 47 behind-the-ear (BTE) hearing aids. A closed-set consonant test was carried out
in both quiet and noise. For the test group, the speech stimuli were low-pass
filtered for each subject at the boundary of the dead region, 0.5 octave above
the boundary, and one octave above the boundary. Cutoff frequencies for the
control group were identical to those of the matched subjects of the test group.
A wide-band condition was included for all subjects. It was found that the two
groups obtained similar scores for testing in quiet and testing in low-level noise.
In other words, for these test conditions, regardless of the presence or absence
of dead regions, both groups of subjects obtained an equal amount of benefit
from amplification. However, testing in high levels of noise found that the group
with suspected dead regions performed more poorly than the control group for
the wide-band condition. While these results differ from those of the Vickers et
al. (2001) study, they do show agreement with the Baer et al. (2002) findings.
In summary, many studies have attempted to investigate
whether high-frequency amplification benefits listeners with
severe-sloping hearing losses. Unfortunately, in many of these
studies it was not possible to provide full audibility of the highfrequency signal. Audibility remains a confounding factor when
determining which hearing-impaired listeners could benefit from
additional high-frequency information. The TEN test was
developed as a means of providing clinicians with a tool to
diagnose dead regions, and thereby a means for deciding when
to provide high-frequency amplification. Chapter 7 reports
speech perception results for 10 subjects with moderate-toprofound high-frequency hearing losses under various low-pass
filter conditions. The confounding factor of audibility was
reduced when measuring speech perception by attempting to
maximize audibility of the speech stimuli. The TEN test was
carried out with all subjects.
- 48 – Chapter 3 The role of audibility in speech perception
- 49 -
Chapter 4
Frequency compression as a means for providing highfrequency information
4.1
Introduction
Chapter 3 established that providing high-frequency audibility to listeners with a
severe high-frequency hearing loss did not always result in positive speech
perception outcomes. Since conventional hearing aids failed to provide the
listener with high-frequency cues, then alternative methods of signal processing
were considered.
Two such methods which were investigated in the current study were:
1. Shifting higher frequencies down to lower frequencies in a hearing aid, and
2. Providing high-frequency information electrically via a cochlear implant.
The first of these two methods is discussed in this chapter and is commonly
referred to as frequency lowering (also known as “frequency shifting” or
“transposition”). The processing aims to preserve useful low-frequency audibility
while providing additional high-frequency information. Past results with
frequency-shifting schemes is summarized in sections 4.2 and 4.3 of the current
chapter. The frequency-shifting scheme selected for the current study is
described in section 4.4 of the current chapter.
4.2
Early attempts at frequency shifting
Various sound-processing schemes have been designed with the aim of
presenting information from high-frequency components of speech at lower
frequencies. These schemes often used non-linear modulation techniques to
shift high-frequency speech components downwards to a lower-frequency
range. The downward shift was typically disproportionate, meaning that
frequency ratios contained in the spectral information were not preserved during
processing. The resulting signals were mixed with those obtained from lower
frequencies. One such scheme was developed by Johansson (1961). The
device consisted of two channels. Frequencies of 150 – 3000 Hz were amplified
- 50 – Chapter 4 Frequency compression as a means for providing high-frequency information
in a conventional manner. Higher frequencies between 4000 – 8000 Hz were
passed through a nonlinear modulator and converted into low-frequency noise
below 1500 Hz. The device was tested on a group of normally-hearing adults
with a simulated hearing loss (Johansson, 1961; Wedenberg, 1961). These
subjects showed improvements in speech perception when listening via the
scheme. However, subsequent studies with hearing-impaired listeners found no
significant improvements in speech intelligibility with the device (Ling, 1968;
Velmans et al., 1983). A major shortcoming of the scheme was that the
processing did not allow for the preservation of significant details of the spectral
shape of the incoming signals.
Alternative schemes have been developed which did present some information
about high-frequency spectral shape. In one such scheme, Velmans (1974)
separated signals into low-pass and high-pass bands, with a crossover
frequency of 4000 Hz. A constant value of 4000 Hz was subtracted from each
frequency present in the high-pass band and the resulting signals were mixed
with those obtained from the low-pass band. Although some positive results
were reported for this scheme for both intelligibility and articulation of speech
(Velmans, 1973; Velmans, 1975), the processing may have had some
disadvantages. For example, some perceptual information may have been lost,
as the frequency ratios in the high-frequency band were not preserved when
shifted to lower frequencies. Similarly to the device tested by Johansson (1961)
it is possible that these earlier schemes may have provided some additional
high-frequency information at the expense of other perceptual cues by
overlapping the shifted and unshifted signals.
Another experimental processing scheme that preserved aspects of the spectral
shape analyzed the high-frequency region of sounds using a bank of band-pass
filters (Posen et al., 1993). The envelopes of signals in these filters were
estimated and used to modulate the amplitudes of an equal number of signal
generators, which produced either pure tones or narrow-band noises at
frequencies lower than those of the corresponding filters. The unmodified lowfrequency signals were combined with the outputs of the signal generators. This
scheme was reported to provide speech intelligibility benefits, but only after
listeners underwent a period of auditory training.
- 51 A simple method of preserving spectral information would be to shift all
frequency components downwards by a constant factor. Although the pitch of
the speech signal would be lowered, the ratios among the frequency
components would be left unchanged by the processing. In particular, the
relationship between the frequencies of the formant peaks in speech would
remain constant. These ratios may be particularly important cues for the
recognition of vowels in speech (Neary, 1989). Some listeners have obtained
speech perception benefits when listening to proportional frequency
compression (Turner et al., 1999). Listeners were required to identify nonsense
syllables spoken by a male and female talker. Improvements in intelligibility for
the female speaker were noted for about half of the subject group. A similar
method of proportional shifting was investigated by McDermott & Dean (2000).
In that study, six subjects with steeply-sloping losses were required to listen to
monosyllabic word lists with and without frequency shifting. The word lists were
lowered in frequency by a factor of 0.6. Despite intensive training of the
subjects, no significant differences were reported in subjects’ abilities to
perceive speech when frequency shifting was enabled.
The perceptual performance of a number of linear and nonlinear frequencycompression schemes was evaluated by Reed et al. (1983). In a preliminary
study, six subjects with normal hearing participated in experiments that
investigated whether any of the schemes could improve the discriminability of
consonant stimuli. Although none of the schemes provided better performance
than a standard condition that applied only low-pass filtering to the stimuli, the
best scheme was found to be a variant that progressively increased the amount
of frequency compression for input frequencies above approximately 1200 Hz.
Lower input frequencies were hardly changed by the processing. Subsequently,
this scheme was one of two variants tested with a small group of hearingimpaired subjects (Reed et al., 1985). None of the three subjects who
completed a consonant identification test with the frequency-compression
scheme obtained higher scores than with linear amplification. Overall, these
findings suggested that, although frequency compression did not provide a
perceptual benefit, the scheme that resulted in the best scores for the
- 52 – Chapter 4 Frequency compression as a means for providing high-frequency information
consonant tests applied increasing amounts of frequency lowering to relatively
high-input frequencies, while leaving lower frequencies unchanged.
4.3
Frequency shifting by means of a “slow-play” method
Proportional frequency shifting, using a “slow-play” method, is an alternative
sound-processing technique that is currently available commercially (Bennett et
al., 1967; McDermott et al., 2001; McDermott et al., 1999). Segments of the
speech signal are recorded and then played back at a slower speed than
employed for recording. The AVR Sonovations company releases commercially
available hearing instruments that incorporate such processing. The latest
products implement digital processing and are known as the NanoXP,
ImpaCtXP, and the LogicomXP. Published data exists about older versions of
the device, known as the TranSonic and ImpaCt DSR675. Incoming signals
dominated by components at frequencies above 2500 Hz are shifted down by a
factor that is programmable for each listener. If the input signal is not dominated
by frequencies above 2500 Hz, then signals are amplified with no frequency
shifting. Positive outcomes were reported when the TranSonic was fitted to a
small number of hearing-impaired children (Davis-Penn et al., 1993;
Rosenhouse, 1990). A similar study with adults found that two of the four
subjects tested demonstrated speech perception benefits with the device
(Parent et al., 1998).
McDermott et al. (1999) reported a study in which five subjects obtained higher
scores with the TranSonic than with their own hearing aids, but suggested that
the amplification characteristics of the TranSonic in the low frequencies, rather
than its frequency shifting characteristics, may have provided most of the
benefit. Only two subjects appeared to obtain additional speech information
specifically from the high-frequency signal components after they were lowered.
A more recent study (McDermott et al., 2001) of the AVR ImpaCt hearing
instrument found little difference in performance between the ImpaCt aid and
the subjects’ own conventional aids. In addition, the subjects’ understanding of
sentences in competing noise was significantly poorer with the ImpaCt.
- 53 -
Speech perception results with a novel frequency-shifting
device
4.4
Generally, studies investigating the perceptual performance resulting from use
of frequency-lowering schemes have provided mixed results. Many novel
processing schemes are possible with the arrival of digital technology and may
be of benefit if appropriate fitting parameters can be found.
The performance of a novel frequency-compression hearing device 1 was
investigated using tests of speech understanding in quiet (Simpson et al.,
2005c). The device consisted of two main parts: a pair of modified behind-theear (BTE) conventional hearing devices, and a Stereo Hearing Aid Research
Processor (SHARP) which was a programmable body-worn device (Zakis et al.,
2004). Sound entered the system through the microphone (or telecoil or direct
audio input) of each BTE, and was processed by the conventional hearing
devices. The BTEs were modified so that their outputs were directed to the
SHARP processor. The SHARP processor then manipulated the signals to
perform frequency compression and sent the outputs to the earphone receivers
located in the BTEs. There was one cable connecting each BTE to the SHARP
processor. Typically, the hearing-aid wearer would fasten the SHARP processor
to a belt or pocket, running the cables to the BTE devices worn on each ear.
The device compressed frequencies above a programmable cut-off, resulting in
those parts of the input signal being shifted to lower frequencies. Below the cutoff, signals were amplified without frequency shifting. Figure 4.1 shows the
input-to-output frequency relationship implemented in the frequencycompression device. Frequency lowering was applied only to signal
components having relatively high frequencies. The frequency compression
was nonlinear and applied progressively larger shifts to components having
increasingly higher frequencies. Consequently, a wide range of high-frequency
input signals resulted in a narrower range of output signals. The possible
advantages of the scheme include no overlap between the shifted and unshifted signals.
1
This work was conducted together with Adam Hersbach and Hugh McDermott. Phonak
Hearing systems provided the hearing devices used in the study.
- 54 – Chapter 4 Frequency compression as a means for providing high-frequency information
10
Cut-Off
Frequency
Output Frequency (kHz)
5
2
1
0.5
0.2
0.1
0.1
0.2
0.5
1
2
5
10
Input Frequency (kHz)
Figure 4.1 An example of the input-frequency to output-frequency relationship
implemented in the experimental frequency-compression hearing device. Figure from
Simpson et al. (2005c).
Low- and mid-frequency information appeared to remain preserved with the
frequency-compression device as all of the first-formant frequency range and
most of the second-formant frequency range were below the cut-off frequency.
A possible disadvantage of the experimental scheme was that it did not
preserve frequency ratios for those high frequencies that were compressed. It is
a possibility that speech perception could be affected if the cut-off frequency
was decreased to lower frequencies, or that the perception of certain sounds,
such as music, may be affected adversely.
Each subject was fitted with identical conventional hearing aids (Phonak Supero
412) using appropriate fitting software (Phonak Fitting Guideline version 8.1).
Speech perception comparisons were made between this hearing device and
the subjects’ own hearing aids (Simpson et al., 2005b). Open-set testing was
carried out using monosyllabic word lists at a level of 65 dBA. Subjects were
selected for the frequency-compression study only if their speech perception
results with the conventional hearing device were approximately the same as,
- 55 or better than, those obtained with their own hearing aids. Each subject had
been wearing the conventional hearing devices for several months prior to the
commencement of the study.
To fit the experimental frequency-compression scheme, each subject’s fitting
program was initially saved into a modified conventional hearing device. A cutoff frequency was chosen individually for each subject based on the audiogram.
As a general rule, the initial cut-off frequency was set to the point at which
conventional amplification provided the listener with minimal audibility. The most
commonly selected cut-off frequencies were 2000 or 2500 Hz where hearing
thresholds were often greater than 90 dB HL. For each subject, the loudness of
the frequency-compressed signals was approximately equalized across
frequency. Subjects were required to wear the frequency-compression device in
place of their conventional instruments for approximately 5 weeks.
Seventeen experienced hearing-aid users with moderate-to-profound
sensorineural hearing loss and sloping audiograms participated in the trial. Two
experiments were carried out. Experiment (a) compared the speech
understanding abilities on a monosyllabic word test of these subjects when
listening via the experimental device with the frequency-compression function
enabled and disabled. Subsequently, experiment (b) investigated whether the
addition of high-frequency audibility without frequency shifting would result in
similar outcomes.
The mean phoneme, consonant, fricative consonants (/f/, /s/, /S/, /v/, /z/, /T/, /j/,
/tS/), and vowel scores obtained for the monosyllabic words test is shown in
Figure 4.2.
Across the 17 subjects, speech recognition scores were compared with the two
hearing-aid processing schemes by means of a two-factor analysis of variance.
A statistically significant improvement for the frequency-compression scheme
(FrC) over the conventional hearing device (CD) was found for phoneme,
consonant, fricative, and vowel scores (p < 0.001). Each subject’s data was
analyzed further with pair-wise comparisons (Holm-Sidak method). Of the 17
subjects, eight obtained a significant (p < 0.05) phoneme score increase with
the experimental frequency-compression scheme, eight subjects showed no
- 56 – Chapter 4 Frequency compression as a means for providing high-frequency information
significant change in scores, and one subject showed a significant score
decrease with the experimental scheme (p = 0.004).
Mean score correct (%)
100
80
CD
FrC
60
40
20
0
Phonemes Consonants
Fricatives
Vowels
Figure 4.2 Mean phoneme, consonant, fricative, and vowel scores obtained by the 17
hearing-impaired subjects who participated in the study when listening to monosyllabic
words. Unfilled columns show mean scores obtained using the conventional hearing
devices (CD), and filled columns show mean scores obtained using the experimental
frequency-compression scheme (FrC). Error bars indicate one standard deviation.
Seven of the eight subjects who performed better with the experimental scheme
participated in Experiment (b). The experimental device was programmed to
provide each subject with additional high-frequency audibility but without
frequency shifting. It was found that the addition of high-frequency gain with no
frequency compression did not result in similar speech perception benefits.
Across all subjects there was no statistically significant difference in speech
scores between the hearing-aid program with additional high-frequency gain
and the conventional hearing device. It seems that the positive outcome
observed for this group of subjects could not have been achieved by increasing
high-frequency audibility alone.
- 57 -
In summary, frequency compression has shown mixed results.
Some studies have reported positive benefits for hearingimpaired
subjects
when
listening
to
frequency-lowering
schemes (Davis-Penn et al., 1993; Parent et al., 1998; Posen et
al., 1993; Rosenhouse, 1990; Simpson et al., 2005c; Turner et
al., 1999; Velmans, 1973; Velmans, 1975), while others have
reported little or no benefit (Ling, 1968; McDermott et al., 2000;
McDermott et al., 2001; Velmans et al., 1983). The large
majority of these studies have been conducted in laboratories
and few have included a take-home trial of the device. The
experimental frequency-compression scheme described by
Simpson et al. (2005c) proved successful in providing additional
high-frequency information for some listeners with moderatelysloping losses whose hearing thresholds became severe-toprofound in the 2000 – 8000 Hz region. The scheme was not
tested with listeners with steeply-sloping hearing losses.
Typically,
these
individuals
have
near-normal
hearing
thresholds in the low frequencies and a profound hearing loss
for frequencies above approximately 1500 Hz. The current
study (described in Chapter 8) examined whether the
experimental frequency-compression scheme described by
Simpson et al. (2005c) would benefit 7 listeners with steeplysloping hearing losses. All subjects were initially fitted with a
conventional device. Speech perception in both quiet and noise
were compared between this device and the frequencycompression device. The APHAB questionnaire (Cox et al.,
1995) was given to subjects at the end of the trial in order to
determine their subjective preferences.
- 58 – Chapter 4 Frequency compression as a means for providing high-frequency information
.
- 59 -
Chapter 5
Cochlear implants
5.1
Introduction
Chapter 4 described altering the signal processing of a hearing device by
compressing high-frequency components of a signal into lower frequencies. In
this technique, an acoustic signal is manipulated and presented to the listener.
This method is one way of attempting to provide the hearing-impaired listener
with additional high-frequency information. A second means of providing this
information is cochlear implantation. A cochlear implant (CI) is a device which is
surgically inserted. An electrode array is placed into the cochlea and used to
electrically stimulate the auditory nerve. Information about sound is thus
transmitted electrically to the user.
This chapter will provide a general overview of how cochlear implants work.
Section 5.2 describes some of the considerations in cochlear implant system
design, particularly the electrode array and the number of channels in the array.
Section 5.3 describes implant systems which are currently available
commercially. Section 5.4 discusses speech-processing strategies with
emphasis on the ACE strategy which was selected for this study.
5.2
Cochlear implant system design
There are a number of CI systems available commercially. A representation of
the complete system is shown in Figure 5.1. All of these systems have the
following in common: a microphone to pick up incoming sound, a signal
processor to convert the incoming sound into electrical signals, a system which
transmits the electrical signal to the electrode array, and an electrode array.
The microphone is located behind the user’s ear. It picks up external sound and
relays it to a speech processor. The speech processor (usually worn behind the
ear) codes the sound electronically. The coded information is then transmitted
via a coil in the form of radio waves across the skin to the stimulator/receiver.
The coil is located behind the ear and is held in place by two magnets, one
- 60 – Chapter 5 Cochlear implants
situated in the centre of the coil and the second placed under the skin. The
stimulator/receiver converts the code into electrical signals which are relayed to
the electrode array. In Figure 5.1 the array has been inserted into the scala
tympani of the cochlea (intracochlear). Electrodes can also be placed outside of
the cochlea (extracochlear), usually on the medial wall or the round window.
Intracochlear electrodes have the advantage of being closer to the neural
elements and therefore, may have finer control of neural firing patterns. The
array consists of several electrode bands arranged in a row which are mounted
on a silicone tube. A wire connects each electrode to the stimulator/receiver.
When the electrodes receive the electrical signal, they deliver a charge to the
auditory nerve fibres.
3
4
1
5
2
Figure 5.1 A cochlear implant system which consists of (1) a microphone to pick up
ambient sound, (2) an external speech processor for coding sound electronically, (3) a
transmitting coil to send the signal as a radio wave across the skin, (4) a
stimulator/receiver which converts the code into electrical signals, and (5) an electrode
array implanted into the cochlea. Figure from Cochlear Ltd.
Although these components are similar between manufacturers, the methods
used in coding sound, transmitting information to the implant, and stimulating
the electrodes differ between devices. Electrode arrays also differ in design.
The arrays vary in length between manufacturers and are inserted into the
- 61 cochlea at depths varying from 10 – 30 mm. Electrode arrays also vary in
number of channels. In single-channel implants, only one electrode was used,
thereby stimulating one fixed site within the cochlea. Some examples of singlechannel electrodes that were implemented clinically were the 3M/House device
(House et al., 1981) and 3M/Vienna implant (Hochmair-Desoyer et al., 1983).
These devices could not make use of the frequency-place coding in the
cochlea. Spectral information was limited, and there were relatively few reported
cases of patients obtaining open-set speech understanding (Gantz et al., 1988).
For this reason multi-channel implants are now used. These have the
advantage of being able to stimulate different places in the cochlea. Electrodes
placed near the apical end of the cochlea stimulate the place along the cochlea
which would normally correspond to low-frequency acoustic signals, whereas
high-frequency information is conveyed by electrodes placed near the basal end
of the cochlea. This is known as “place-coding” as the implant attempts to mimic
the tonotopic organization in the cochlea.
For place-coding to be effective, the current must be localized enough to be
able deliver charge to separate groups of auditory nerve fibres. The type of
electrode configuration which delivers current to the neurons is referred to as
the mode of stimulation. There are currently three main modes of stimulation in
use. A representation of these three modes is shown in Figure 5.2.
Common-ground
Bipolar
Monopolar
Figure 5.2 Representation of the three types of electrode configurations used in cochlear
implant systems. In common-ground stimulation (shown by the left illustration), current
passes between the active electrode and all other electrodes in the array. The middle
illustration shows bipolar stimulation in which electric current is passed between two
electrodes in close proximity. Monopolar stimulation is shown in the right illustration. In
this stimulation mode electric current is passed between an active electrode and a
remote ground electrode.
- 62 – Chapter 5 Cochlear implants
In monopolar (MP) stimulation, the electric current is passed between an active
electrode and a reference electrode which is situated outside the cochlea. In
bipolar (BP) stimulation, the active electrode is placed close to the reference
electrode. In common- ground (CG) stimulation, current is passed between the
active electrode and all other electrodes in the array, which are connected
together and serve as a common return path for the current. Physiological
measures have shown that monopolar stimulation results in a larger spread of
excitation when compared to bipolar stimulation (Bierer et al., 2002; van den
Honert et al., 1987). However, speech recognition with monopolar stimulation
has been found to be equivalent or better than that with bipolar stimulation
(Kileny et al., 1998; Pfingst et al., 1997; Pfingst et al., 2001). In addition, the
amount of current needed to elicit a hearing perception is lower for monopolar
than bipolar stimulation. This minimises the amount of power required to drive
the device. For these reasons, monopolar stimulation is used most commonly in
modern CI systems.
Section 5.2 has given a general overview of the CI system. The following
section describes the CI systems which were commercially available at the time
of writing.
5.3
Current implant systems commercially available
There are currently four main manufacturers of implant systems: Advanced
Bionics Corporation (United States of America), Med-El (Austria), MXM
(France) and Cochlear Ltd. (Australia).
Advanced Bionics’ latest product at the time of writing is the “HiResolution”
implant system which consists of the HiRes 90K Bionic Ear Implant and
electrode array, and HiRes AURIA BTE speech processor (Koch et al., 2004).
The array is pre-coiled with 16 electrodes which terminate in square contacts.
The stimulator/receiver and coil are housed in a titanium case. The device is
able to stimulate at a stimulation rate (the number of electrical current pulses
per second that can be presented across the electrode array) of 83 000
pulses/s.
The PULSARci and COMBI 40+ implant systems (Helms et al., 1997) are
produced by Med-El. The electrode array has a soft electrode carrier designed
- 63 to allow for deep insertion of up to 30 mm. The electrodes are spaced 2.8 mm
apart over 26.4 mm, a larger distance than other commercial arrays. There are
an additional two types of arrays for use in cases of cochlear ossification. The
stimulator/receiver and coil are housed in a ceramic case. The device can
stimulate at 18 000 pulses/s across the electrodes. The devices produced by
Advanced Bionics and Med-El all allow for a number of different stimulation
modes and processing strategies. The Digisonic SP is a multi-channel CI
produced by MXM. The electrode array has 15 electrodes spaced over a 24 mm
length. Stimulation occurs in common-ground mode. The stimulator/receiver is
encased in a ceramic convex casing. The device is capable of providing a
stimulation rate of 24 000 pulses/s across electrodes.
For the study described in Chapter 9, the Nucleus Freedom system (Cochlear
Ltd., Australia) was used. It is shown in Figure 5.3. It is a multi-channel
intracochlear device consisting of an ear-level microphone, a BTE speech
processor, and an implanted stimulator/receiver connected to an electrode array
with 22 electrodes. The electrodes are spaced over a distance of 15 mm. The
stimulator/receiver is encased in titanium. When programming, the clinician has
a choice of several speech processing and stimulation strategies. The device
allows for stimulation rates of up to 31 500 pulses/s across electrodes.
2
4
3
1
Figure 5.3 The Nucleus Freedom CI system which consists of the (1) Freedom BTE
speech processor, (2) transmitting coil, (3) contour advance electrode array, and (4) the
stimulator/receiver. Figure from Cochlear Ltd.
- 64 – Chapter 5 Cochlear implants
5.4
Speech-coding strategies
Implant systems differ in the signal-processing strategy implemented for
converting the sound signals into electrical stimuli. Because of the large number
of processing strategies currently in use, it is beyond the scope of this thesis to
discuss each one in detail. The more common strategies such as Simultaneous
Analog Stimulation (SAS), Continuous Interleaved Sampling (CIS), Spectral
Peak (SPEAK) and High Resolution (HiRes) will be described as well as a
detailed description of the ACE strategy which was chosen for this study. For a
more comprehensive review of additional signal processing strategies, see
Loizou (1998).
Despite differing in the approach to coding speech, the strategies listed above
are all able to provide the average user with reasonable speech recognition, at
least in quiet. Table 5.1 shows published monosyllabic word scores for ACE,
CIS, and HiRes. Although some caution should be taken when comparing
results across studies, no one strategy appears to perform considerably better
than the other two schemes. Scores appear similar if subjects had an equal
amount of listening experience with the CI device. With 12 months’ listening
experience, the mean monosyllabic word score with the ACE strategy was
55.2% (S.D. 11.9) (Skinner et al., 2002) and 55.1% (S.D. 20) (Helms et al.,
1997) with the CIS strategy. ACE was selected for the current study as it is the
most common clinically selected strategy in the Melbourne Cochlear Implant
Clinic and published data at the time of writing showed no clear advantage in
selecting another scheme.
In order to understand how speech is coded by the implant, one must first have
an understanding of speech as an acoustic signal. A single unit of sound is
referred to as a phoneme. Each phoneme is unique in terms of frequency
content and intensity. Speech consists of a rapid succession of phonemes. As
such, speech is constantly changing in intensity and frequency. Changes in
intensity over time are known as temporal cues, while changes in the
distribution of energy across frequency are referred to as spectral cues.
Speech strategies can be divided into two approaches: the filterbank approach
and the waveform approach. All the strategies mentioned above with the
exception of SAS use the filterbank approach.
- 65 -
Strategy
Study
Subjects
Experience
(months)
ACE
CIS
SD
Kiefer et al. (2001)
11
>1
49.8
30.9
David et al. (2003)
26
12
55.2
11.9
Skinner et al. (2002)
12
1.5
37.9
14.9
Helms et al. (1997)
55
1
36.2
22.8
3
44.7
23.2
6
49.7
20.2
12
55.1
20.0
Muller et al. (2002)
9
>2
43.1
18.8
Ziese et al. (2000)
6
>3
50.9
24
Hamzavi et al.
(2003)
21
12
40.6
9.6
41.9
7.5
1
24.5
16.3
6
37.3
17.5
12
44.6
19.9
3
50.0
25.1
25
Gstoettner et al.
(1998)
HiRes
Mean
monosyllabic
word score (%)
Koch et al. (2004)
21
51
Table 5.1 Summary of published studies reporting monosyllabic word recognition using
current CI speech-processing strategies. Table reproduced from Wilson (2006).
Electric information is conveyed by a series of narrow biphasic pulses. Each
pulse consists of two short time intervals. In each of these intervals, a charge of
equal magnitude, but flowing in opposite directions, is delivered to the active
electrode. These pulses can be varied in terms of amplitude, duration, and
onset time. This is known as pulsatile stimulation. A representation of pulsatile
stimulation is shown in Figure 5.4. Loudness perception by the user can be
changed by varying the amount of charge. The recipient indicates at what
current level he/she can just hear the stimulus (threshold or T level) as well as
what current level produces an uncomfortably loud sensation (comfort or C
- 66 – Chapter 5 Cochlear implants
level). The procedure of measuring T and C levels is referred to as “mapping”.
The range between the T and C levels is known as the recipient’s electrical
dynamic range. Speech signals, after processing as described later, can then
be mapped into this electrical dynamic range.
In SAS (Kessler, 1999) the incoming signal is analysed in terms of frequency
and amplitude and divided into partially overlapping filter bands. Each filtered
waveform is compressed to fit the patient’s dynamic range and delivered
simultaneously as continuously varying currents to the active electrodes. The
acoustic waveform is thus represented electrically and sent to the electrode.
This is known as analog stimulation. A representation of analog stimulation is
shown in Figure 5.4.
Figure 5.4 The two types of stimulation waveform used in implant systems. Time is
represented on the x-axis, and current is represented on the y-axis. The upper panel
shows pulsatile stimulation, while the lower panel shows analog stimulation. Figure
reproduced from McDermott (2004).
Whereas SAS uses the waveform at the output of each filter, filterbank schemes
attempt to convey spectral information by making use of the signal envelope.
The pulse rate applied to the electrical stimulation is independent of any
characteristic of the input signal. Examples of these strategies include CIS,
ACE, and SPEAK. In CIS (Wilson et al., 1991) the incoming signal is passed
through a bank of bandpass filters. The envelopes of the waveform are
extracted, compressed and used to control the amplitude of the biphasic pulses.
In original implementations of the CIS scheme, balanced biphasic pulse trains
were delivered to 6-8 electrodes at a constant rate (800 Hz or more) in a non-
- 67 overlapping fashion. Newer implementations of the scheme, such as HiRes,
allow for stimulation on up to 16 electrodes as well as higher stimulation rates of
up to 90 000 pulses/s across electrodes.
The SPEAK (McDermott et al., 1993; Seligman et al., 1995) and ACE (Vandali
et al., 2000a) strategies are functionally similar but use different rates with the
ACE strategy allowing for higher rates. The ACE strategy is described below.
5.4.1 The Advanced Combination Encoder (ACE) speech-
processing strategy
The ACE (Vandali et al., 2000a) strategy was selected for this study and is
represented in Figure 5.5. The waveform is picked up by the microphone of the
speech processor. The signal then undergoes “front-end” processing which
refers to the processing of the signal that occurs before the spectrum is
estimated by the filterbank. This includes a pre-amplifier, pre-emphasis, and
automatic gain control (AGC). The signal from the microphone is increased, or
pre-amplified, to a level which can be easily manipulated by the rest of the
signal processing. In addition, the processor applies more amplification (or gain)
to higher frequencies.
This “pre-emphasis” of the signal results in a level increase of 6 dB for each
octave in frequency. This allows for relatively weak consonants, which have a
predominately high-frequency content, to be able to be heard at approximately
equivalent levels of vowels. Vowels sounds have a predominately lowfrequency content and are usually more intense than consonant sounds. A
single input level, or gain, would be inappropriate in all situations. For example,
the microphone is positioned only a short distance from the user’s mouth. The
level of the user’s own voice would thus usually be higher than that of the
person they are engaged in conversation with. In this case, the user’s voice
requires less amplification, whereas the voice of the person they are talking to
requires a larger amount of amplification. The processor attempts to
compensate for this by changing the gain automatically by means of the AGC.
This can also be achieved manually by the listener with the sensitivity control.
The sensitivity control is initially set to an optimum setting for listening to speech
at conversational levels. It determines when the AGC becomes active.
- 68 – Chapter 5 Cochlear implants
Speech waveform
Front end
processing
Filterbank
Envelope
Selection of
maxima
C
Conversion to
patient’s T and
C levels
T
Pulse generator
Electrodes stimulated non-simultaneously
Figure 5.5 A representation of the ACE speech-processing strategy. The incoming signal
first undergoes front-end processing before the short-term spectrum is estimated by a
number of bandpass filters. The filters with the largest outputs (maxima) are periodically
selected for further analysis. The amplitudes of the maxima are converted into the
listener’s dynamic range (delimited by the T and C levels) and transmitted to the
corresponding electrodes.
The AGC changes the gain automatically dependent on the amplitude of the
incoming signal. For intense inputs, like the user’s own voice, the processor
- 69 reduces the gain. As the amplitude of the signal is reduced, the gain is gradually
brought back to its original value.
Once the signal has undergone front-end processing, it is analysed in terms of
frequency and amplitude and divided into partially overlapping filter bands by
means of a Fast Fourier Transform (FFT). The number of filter bands is
dependent on the number of useable electrodes, usually 20 or 22. The
Electrode
frequency-to-electrode allocation is shown in Figure 5.6.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
100
200
500
5000
10000
Frequency (Hz)
Figure 5.6 Frequency allocation for the current default clinical filter bank used in
Melbourne, Australia with the Freedom device when the ACE strategy is selected. The
frequencies assigned to each electrode can be seen for frequency (Hz) on the x-axis
plotted against electrode number on the y-axis.
Typically, the filter bands are linearly spaced from 188-1312 Hz, and
logarithmically spaced thereafter up to 7938 Hz. The envelope from each filter
band is extracted and a subset of filters with the largest outputs is periodically
selected for further analysis. These are referred to as “maxima” and can range
in number from 1-20, although the more typical number of maxima ranges from
6-10.
- 70 – Chapter 5 Cochlear implants
The amplitudes of these maxima are converted via an amplitude conversion
function whereby the acoustic levels are converted into electrical levels. The
upper and lower limits of this function are referred to as the input dynamic range
(IDR). IDR is illustrated by the example in Figure 5.7.
Current level
C
T
0
70
25
100
45 dB IDR
Input level (dB SPL)
Figure 5.7 An example of the amplitude conversion function for an IDR of 45 dB. Input
level (dB SPL) is shown on the x-axis, and current level is shown on the y-axis. The solid
black line represents the amplitude conversion function. The dotted red lines show how
this function would change if the sensitivity setting was altered. Higher input levels are
picked up by the speech processor by turning the sensitivity setting down. Lower input
signals are picked up by the processor by turning the sensitivity setting up.
This range is 45 dB in the Freedom processor, but can be adjusted. For
technical reasons the IDR varies between different speech processors. This
function together with the patient’s measured threshold (T) and comfortable
loudness (C) levels determines the level of the electric stimuli. The input level of
the signal is plotted against current level for an IDR of 45 dB.
The solid black line represents the amplitude conversion function. In this case
the sensitivity control has been set to position which would be appropriate for a
70 dB SPL maximum input level. Therefore, an input level of 70 dB SPL will
- 71 result in an electrical stimulus level close to the listener’s C level. This is the
highest level at which stimulation will occur. An input level which is higher than
70 dB SPL will still correspond to the listener’s C level. An input level of 25 dB
SPL will stimulate the listener at T level. Any inputs lower than 25 dB SPL will
not be processed. The dotted red lines illustrate how this function would change
if the sensitivity control was turned up or down. By turning the sensitivity down,
input levels would need to be higher in order to be picked up by the processor.
Similarly, lower input levels are processed by turning the sensitivity up.
The signals are then transmitted to the corresponding electrodes. Biphasic
pulses are delivered to the electrodes in a non-simultaneous sequence at a
constant rate. In other words, only one electrode ever delivers current at one
time. The stimulation rate can be chosen from a range of 250 – 3500 pulses/s
per electrode. A moderate stimulation rate of 900 pulses/s per electrode was
selected for the current study. This rate is the most commonly clinically selected
rate with the ACE strategy in the Melbourne Cochlear Implant Clinic. Additional
support for the selection of a moderate rate was reported by Vandali et al.
(2000b). The effects of different stimulation rates were investigated in 6 adult CI
users. Rates of 250, 807, and 1615 pulses/s per electrode were compared with
subjects using a processing strategy similar to that of the ACE scheme.
Subjects were given take-home experience with each stimulation rate. Tests on
monosyllabic word understanding in quiet and sentence testing in noise were
carried out. No significant differences in speech recognition across the group
were observed between rates of 250 and 807 pulses/s. Significantly poorer
performance was observed for some tests for the 1615 pulse/s rate.
In summary, cochlear implants have become widely recognized
as a safe and effective means of treating deafness, providing
improvements in speech understanding for many users. For the
study described in Chapter 9, the Nucleus Freedom system
(Cochlear Ltd) was selected with the ACE speech-processing
strategy with 8 maxima and a stimulation rate of 900 pulses/s
per electrode.
- 72 – Chapter 5 Cochlear implants
- 73 -
Chapter 6
Cochlear implantation and the benefits of residual hearing
6.1
Introduction
When cochlear implantation was first introduced, it was considered suitable only
for those individuals with bilateral-total hearing losses. As cochlear implants
have provided very successful outcomes for the large majority of these
listeners, the selection criteria for implantation have been relaxed to include
individuals with more residual hearing (Dowell et al., 2004; Rubinstein et al.,
1999). Listeners who have a severe loss in the high frequencies but who have
good hearing sensitivity in the low frequencies often find conventional hearing
aids of little use. The option of cochlear implantation is now being explored with
these patients. This chapter outlines current literature available on implantation
for patients with steeply-sloping losses. Hearing preservation and speech
perception outcomes will be described in section 6.3. In addition, mapping
considerations for these patients, and in particular, frequency-to-electrode
allocations will be discussed in section 6.2.
Electrode insertion depth and frequency-place mismatch
in the cochlea
6.2
This section aims to link the following concepts: electrode insertion depth and
position, the Greenwood function, and the site of the stimulated neurons. As
discussed in Chapter 2 (see section 2.3), in the normally-hearing ear, the place
of maximum displacement of the basilar membrane is dependent on the
frequency content of the signal. Auditory nerve fibres have also been found to
respond maximally to a characteristic frequency. Cochlear implants take
advantage of this characteristic when providing electrical stimulation.
Stimulation with electrodes placed closer to the apex of the cochlea will
generally result in a lower pitch sensation than with electrodes placed closer to
the base. This is supported by studies which have shown that subjective
estimates of pitch with electrical stimulation generally decreases monotonically
as insertion depth increases (Cohen et al., 1996; Deman et al., 2004).
- 74 – Chapter 6 Cochlear implantation and the benefits of residual hearing
Electrode insertion depth in the cochlea can be described in one of two ways:
angle and distance. Insertion-depth distance refers to the distance in mm the
electrode array is inserted from a fixed reference point. By knowing the depth of
insertion into the cochlea, one can make an estimate of the acoustic frequency
range the array’s placement would correspond with. The average insertion
depth has been estimated to be up to 20 mm for Nucleus electrode arrays
(Ketten et al., 1998) and as deep as 30 mm (Gstoettner et al., 1999) for Med-El
arrays.
The most commonly used function to calculate this is the Greenwood (1990)
equation which describes the characteristic frequency along the Organ of Corti
as a function of cochlear place. For example, using Greenwood’s equation, a
frequency range of 1000 -12 000 Hz could be assumed for a 20 mm insertion
depth, assuming that the average cochlea has a length of 35 mm (Greenwood,
1990; Ketten et al., 1998; Ulehlova et al., 1987).
Alternatively, insertion depth can be described in terms of angle (in degrees).
To determine insertion angle, a radiographic image of the electrode array is
taken post-operatively. The skull is placed in a certain position, known as a
modified Stenvers view (Marsh et al., 1993; Xu et al., 2000). Fixed reference
points are used to define the 0º orientation. One complete turn in the cochlea
from this reference point thus corresponds to 360º. The relationship between
angle and percentage length along the Organ of Corti has been shown by
Bredberg (1968). This relationship together with the Greenwood equation
enables insertion angles to be used to estimate characteristic frequencies
corresponding to the electrodes on the array.
If normal acoustic hearing is mimicked as closely as possible, we could in
practice assign frequencies in the speech processor corresponding to the
acoustic frequency range where the electrode array is estimated to be placed.
In effect, this assignment of frequencies could be seen to be in tonotopic
agreement with the normal ear. Such a map shall be referred to as a “placematched” map. An assignment of frequencies in the speech processor that was
different from this acoustic range would result in a map which was tonotopically
shifted. In other words, the frequency range assigned to the stimulation region
- 75 in the cochlea can be very different from the frequency range of the place in the
cochlea that normally responds to the same frequencies acoustically.
The effects of tonotopic shifting have been investigated in several studies. The
effect of insertion depth was examined with five normally-hearing listeners
(Dorman et al., 1997). A simulation of a five-channel implant was created with
electrodes separated by 4 mm. This was achieved by a noise vocoder similar to
the one represented by Figure 6.1.
Speech waveform
Analysis
Filters
Envelope
extraction
Modulate
noise source
by envelopes
Output
Filters
Figure 6.1 Block diagram representing signal processing in a noise vocoder simulation.
The signal was divided into five bandpass analysis filters. The signal envelope
was extracted from each filter. An acoustic source, such as noise, was
- 76 – Chapter 6 Cochlear implantation and the benefits of residual hearing
modulated by the envelopes. The level of the noise in each filter was
determined by the level of the signal envelope in the corresponding analysis
filter. The modulated noise bands were then summed and presented to the
listener. For this experiment, the centre frequencies for each output band were
altered to create different simulated insertion depths. The centre frequencies of
the analysis bands were 418, 748, 1330, 2390, and 4280 Hz respectively. In the
baseline condition, the centre frequency of the analysis filters and that of the
output filters were matched exactly. For all other conditions, the output signal
was shifted upwards in frequency to simulate the effects of different insertion
depths. To simulate a 25 mm insertion depth, the output channels had a centre
frequency range of 493 – 5840 Hz. To simulate a 24 mm insertion depth, the
output channels had a centre frequency range of 590 - 6729 Hz. To simulate a
23 mm insertion depth, the output channels had a centre frequency range of
703 - 7751 Hz. To simulate a 22 mm insertion depth, the output channels had a
centre frequency range of 831 - 8924 Hz. Subjects were required to identify
vowels, consonants, and sentences for each simulated insertion depth as well
as the baseline condition. It was found that insertion depth had a significant
effect on results. Performance at 22 and 23 mm was poorer than the baseline
condition for vowel, consonant and sentence recognition. Performance at 25
mm did not differ from the baseline condition. The greater the upward frequency
shift away from the baseline condition, the poorer subjects performed.
In the experiment described above, output channels were shifted upwards in
frequency. It was found that performance was affected negatively by these
frequency shifts. In cochlear implants, clinical mapping is seldom based on the
input acoustic frequency range. Rather, the common approach is to assign the
largest acoustic range of frequency possible, regardless of the individual
patient’s electrode insertion depth. For example, the electrode array’s
placement in the cochlea may correspond to an acoustic frequency range of
500 – 6000 Hz. However, the frequency range assigned in the speech
processor would often be 125 – 8000 Hz. This results in a compression of
frequency-place information as a wider acoustic range has been assigned in the
speech processor compared to the cochlear region where the electrode array
was placed.
- 77 Fu & Shannon (1999) reported on five CI listeners who were required to listen to
vowels that had been frequency-shifted. The best performance was found at the
frequency allocation that was the closet match to their clinical frequency
allocation. Subjects had been wearing their clinical programs for at least 6
months prior to the study taking place. The results implied that recipients were
able to adapt to an electric frequency allocation which was likely to have been
tonotopically shifted.
The effects of frequency-place compression and expansion were investigated
with six normal-hearing listeners who completed speech recognition tasks when
listening to simulations created by a noise-band vocoder (Baskent et al., 2003).
In the frequency-place matched condition, or baseline, the center frequencies of
the analysis and output channels were matched. For all other conditions, the
analysis frequency bandwidth was expanded or compressed. The frequency
ranges of these conditions are represented in Figure 6.2. No training was given
to subjects. Speech recognition ability deteriorated for all conditions of
frequency expansion and compression. The matched condition always provided
the best results, even when this condition eliminated some acoustic information.
Frequency-place compression or expansion effects were further investigated
with six implant users in a study which manipulated implant signal processing
by changing the frequency range assigned to the electrode array (Baskent et
al., 2004). A frequency-place matched map was created for the matched
condition. The frequency range assigned to the speech processor was the
same as the frequency range estimated to correspond to the electrode array’s
position in the cochlea. This was estimated according to the Greenwood
function (Greenwood, 1990). The input acoustic range was manipulated in the
experimental conditions by frequency shifting, compression, or expansion. The
results agreed with those found for normal-hearing subjects when listening to
the noise vocoder. Speech recognition was affected negatively by changes in
the frequency-to-electrode allocation with the best speech scores being
obtained with frequency-place maps that provided similar information to the
matched condition.
A further study conducted by Baskent & Shannon (2005) tested variations in
frequency-place and insertion depth. Four CI subjects with Med-El Combi 40+
- 78 – Chapter 6 Cochlear implantation and the benefits of residual hearing
devices participated in the study. An average insertion depth of 31 mm was
assumed for all subjects, as well as an average cochlear length of 35 mm. The
10 middle electrodes of the array were selected for the baseline condition. The
frequency allocation for these electrodes was selected to place-match
frequencies in the cochlea. This frequency range was calculated in a similar
manner to that described in previous experiments where the acoustic frequency
range was estimated according to the Greenwood equation. The effects of
insertion depth were investigated by setting the input frequency range of the
processor the same as that of the baseline condition, but disabling apical
electrodes. This resulted in a large amount of acoustic information being
compressed into a smaller cochlear region.
Input range
Compressed
0.5
Matched
Expanded
1
2
4
8
16
32
Acoustic frequency range (kHz)
Figure 6.2 Schematic diagram representing the effects of frequency-place expansion and
compression investigated by Baskent & Shannon (2003). The matched condition is
shown by the solid black filled area. For this condition, the frequency range of the input
and output filters were identical. In conditions of expansion or compression the
frequency range of the analysis bands differed from that of the output bands. A larger
input frequency range was mapped onto a smaller output range in conditions of
compression, whereas a smaller input frequency range was mapped onto a larger output
range in conditions of expansion.
The second part of the experiment adjusted the input frequency range to placematch the selected insertion depth. For example, if only three electrodes were
activated and these corresponded to a cochlear region of 3000 – 6000 Hz, then
- 79 a frequency range of 3000 – 6000 Hz was assigned in the speech processor.
Subjects were required to complete speech perception tasks for each condition.
In general, it was found that speech perception performance decreased as
insertion depth became shallower. For full insertion conditions, maps that were
matched in frequency-place provided the highest speech scores. However, for
shallow insertions the matched map did not provide the best performance. In
these cases, the matched map eliminated too much low-frequency acoustic
information making the speech signal largely unintelligible.
The assumption could be made from these studies that speech perception may
be negatively affected when a tonotopic shift occurs. However, this doesn’t
necessarily appear to be the case. It is difficult to separate the effect of insertion
depth from other factors, such as neural survival. In addition, in many of these
experiments the number of activated electrodes was not held constant across
conditions. Speech perception may have been affected by both the frequency
range as well as the number of available electrodes. Finally, the participants
would have been more familiar with the sound quality of their own maps and
may have required more time to adjust to an alternative program.
There is some evidence to suggest that CI recipients may require substantial
periods of time to compensate for frequency-place mismatches. Three adults
using the Nucleus-22 implant device were required to listen to an experimental
map for 3 months which differed from their own clinical maps by a tonotopic
shift of 2-4 mm (Fu et al., 2002). Speech scores for the experimental map
remained significantly lower at the 3-month point when compared to the clinical
map. However, some of the speech recognition testing showed a gradual
improvement in performance with the experimental map over time.
Upward frequency shifts were investigated for one CI recipient in tests of
sentence understanding in noise, and consonant and vowel recognition in quiet
(Dorman et al., 2003). The frequency representation of the subject’s program
was shifted by 3.2 mm and 6.9 mm. The 3.2 mm shift resulted in a small
decrease in scores, which largely recovered after 1 week of use. The 6.9 mm
shift resulted in much larger decreases in scores, which did not recover with
experience. Vowel scores were particularly affected by the shift.
- 80 – Chapter 6 Cochlear implantation and the benefits of residual hearing
Normal-hearing listeners participated in training exercises to determine whether
they could adapt to speech that had been spectrally shifted (Fu et al., 2003;
Rosen et al., 1999). Subjects participated in listening to noise vocoder
simulations that were systematically shifted upwards in frequency. Speech
recognition scores deteriorated after initial experience. However after training,
speech scores improved to close to baseline scores. For the Fu et al. (2003)
study improvements were restricted to conditions that had been trained, and did
not generalize to other spectral shifts where no training had taken place.
Results suggested that listeners were able to accommodate for alternate
speech patterns with training experience, at least in the short-term.
A common assumption of the studies described above is that Greenwood’s
function accurately converts frequency range to cochlear location. While this
mapping function may be accurate for the basilar membrane, it may not hold
true for stimulation occurring at the spiral ganglion level. The explanation is
linked to the anatomy of the inner ear. The spiral ganglion cells in the modiolus
extend around 1.875 turns as compared to 2.625 turns for the Organ of Corti
(Kawano et al., 1996). Thus, spiral ganglion cells in the apical end of the
cochlea normally responding to a given acoustic frequency may be situated in a
position that is more basal than the point of maximum excitation of the basilar
membrane. This finding is further confirmed by a study in which the length of
the Organ of Corti and spiral ganglion cells were measured in nine human
cadaver cochleae (Leake et al., 2005). It was found that the mean spiral
ganglion length was 13.69 mm. This was far shorter than the mean Organ of
Corti length which was measured to be 33.13 mm. A function is currently being
developed to calculate frequency at the spiral ganglion level by adding a
correction factor into Greenwood’s equation.
Psychophysical evidence supporting anatomical studies that Greenwood’s
function may not be accurate at the spiral ganglion level was described in a
study in which thirteen CI recipients participated, all of whom had some residual
hearing in the non-implanted ear (Blamey et al., 1996). Alternating electric pulse
trains and pure tones were presented to subjects. The acoustic frequency and
electric pulse rate were held constant, while the electrode position was varied.
Subjects were asked to indicate which of the two sounds was “higher” in pitch.
- 81 Despite a large degree of variability among subjects’ responses, there was a
consistent trend for the electrical stimuli that matched the pure tones to be more
basal than expected from the characteristic frequency coordinates of the basilar
membrane.
In summary, some evidence exists for the choice of a matched frequency-toelectrode place allocation (Baskent et al., 2003; Baskent et al., 2004; Baskent et
al., 2005; Dorman et al., 1997; Friesen et al., 2001; Fu et al., 1999). A map
which mimics the normal ear in terms of frequency-place may result in the best
outcome for patients in terms of speech perception as well as the time taken to
adjust to the implant. These findings have important implications for cochlear
implant mapping. Historically, most traditional CI patients had a bilateral severeor-profound hearing loss. As such, it was desirable to provide the patient with as
much information about the speech signal as possible. Thus, the frequency
range assigned in the speech processor was as broad as possible. However, as
more patients with residual low-frequency hearing are implanted, it allows for
the possibility of place-matched mapping. This is because the listener’s residual
hearing is able to provide information about low-frequency components of the
signal. Speech perception may be improved if information could be presented
closer to the normal frequency-place in the cochlea.
The following section discusses studies which have investigated the use of
cochlear implants together with hearing aids in CI recipients with residual
hearing in one or both ears post-operatively.
6.3
Combining electric and acoustic stimulation
As the selection criteria for implants were relaxed, more patients were seen with
some residual hearing in the non-implanted (contralateral) ear. At first, it was
unclear whether a patient should continue to wear a hearing aid in the
contralateral ear. The main concern was whether it would be possible for the
auditory system to integrate the electric stimulation from the implant (CI)
together with the acoustic stimulation from the hearing aid (HA). The
combination of a cochlear implant in one ear and acoustic hearing is known as
“bimodal” hearing. The acoustic hearing could be from the contralateral ear,
implanted ear, or both ears. A bimodal advantage is when there is greater
- 82 – Chapter 6 Cochlear implantation and the benefits of residual hearing
perceptual benefit from the combined signal of the implant together with the
acoustic hearing (CIHA), than when compared with wearing the CI or HA alone.
6.3.1 Combining electric and acoustic stimulation in opposite
ears
For brevity, a summary of studies which have investigated the use of cochlear
implants together with a hearing aid in the opposite ear is shown in Table 6.1. In
general, bimodal benefits have been reported in both adults (Armstrong et al.,
1997; Ching et al., 2004; Dooley et al., 1993; Hamzavi et al., 2004; Mok et al.,
2006; Shallop et al., 1992) and children (Ching et al., 2001; Dettman et al.,
2004). These benefits have included improved speech perception in both quiet
and noise, as well as improved localization ability and sound quality. However,
there have been cases where a small number of individuals have shown poorer
performance from wearing a hearing aid together with an implant when
compared to wearing the implant alone (Hamzavi et al., 2004; Mok et al., 2006).
For example, Mok et al. (2006) carried out speech perception testing with 14 CI
recipients. Scores were compared across three conditions: HA alone, CI alone,
and CIHA. For CNC word testing in quiet, three of the subjects obtained
significantly higher scores for the CIHA condition. Two subjects obtained
significantly poorer performance for the CIHA condition when compared to CI
alone for word testing in quiet. The remaining 9 subjects showed no significant
difference between CI alone and CIHA scores. Further analysis on word,
phoneme and consonant scores found a strong correlation between aided
thresholds and bimodal benefits. Subjects with poorer aided thresholds at 1000
and 2000 Hz showed greater improvements in the CIHA condition. For aided
thresholds at 4000 Hz, subjects were divided into two groups: those with
thresholds within the audiometer limit of 80 dB HL and those where no aided
threshold could be measured. Subjects who had no measurable aided threshold
at 4000 Hz were found to have significantly greater CIHA scores. No significant
correlation was found between the low-frequency aided thresholds of 250 and
500 Hz and CIHA scores.
- 83 AUTHORS
YEAR
PUBLISHED
NUMBER
OF
SUBJECTS
METHOD
OUTCOMES
Armstrong et
al.
1997
12 adults
CI alone and CIHA were
compared in tests of
speech perception in quiet
and noise.
Ching et al.
2001
16 children
Ching et al.
2004
21 adults
Dettman et
al.
2004
16 children
Hamzavi et
al.
2004
7 adults
HA alone, CI alone, and
CIHA were compared in
tests of speech perception
in quiet and localization.
HA alone, CI alone and
CIHA were compared in
tests of speech perception
in noise, and localization.
Pre-operative scores with
binaural HAs were
compared with CI alone
and CIHA post-operative
scores in tests of speech
perception in quiet.
HA alone, CI alone, and
CIHA were compared in
tests of speech perception
in quiet.
Mean group results showed
higher scores for CIHA than
CI alone. A larger benefit for
the CIHA condition was
shown for testing in noise.
All subjects showed a benefit
for the bimodal condition in at
least one of the tests carried
out.
All subjects showed a benefit
for the bimodal condition in at
least one of the tests carried
out.
A significant bimodal benefit
was found for the group.
Mok et al.
2006
14 adults
HA alone, CI alone, and
CIHA were compared in
tests of speech perception
in quiet and noise.
Shallop et
al.
1992
7 adults
Tyler et al.
2002
3 adults
Pre-operative scores with
binaural HAs were
compared with HA alone,
CI alone, and CIHA postoperative scores in tests of
speech perception in quiet.
CI alone and CIHA were
compared in tests of
speech perception in quiet
and noise and localization.
On average, for sentences,
the group scored 79% with CI
alone and 88% with CIHA.
One subject performed worse
in the CIHA condition in one
of the three tests carried out.
9 subjects showed a benefit
for the bimodal condition in at
least one of the tests carried
out. 2 subjects obtained
poorer performance in the
CIHA condition.
The pre-operative mean
score for sentences across
the group was 20%. At six
months post-operatively, this
improved to 37% with CI
alone, and 57% with CIHA.
2 subjects showed a benefit
for the bimodal condition in at
least one of the tests carried
out. One subject obtained
poorer performance in the
CIHA condition.
Table 6.1 Summary of studies investigating the benefits of wearing a CI together with a
HA in children and adults. Abbreviations for conditions are shown as hearing aid alone
(HA alone), cochlear implant alone (CI alone), and cochlear implant together with a
hearing aid (CIHA).
It appeared that subjects with worse hearing for the mid-and high frequencies in
the contralateral ear obtained a greater bimodal advantage. All subjects were
programmed with standard maps, typically providing an acoustic frequency
- 84 – Chapter 6 Cochlear implantation and the benefits of residual hearing
range of 125 – 8000 Hz. Thus, the frequency information from the CI had been
both compressed and tonotopically shifted. The HA provided a similar frequency
range, but the signal was provided acoustically, rather than electrically. Thus,
the signal was presented at the normal place along the cochlea. There may
have been a mismatch in perception of frequency between a signal presented
to the HA and an identical signal presented to the CI. The results imply that the
signal from the hearing aid may have interfered with signals from the cochlear
implant at mid-high frequencies. It would have been interesting to see whether
patients’ scores would have been affected if either the HA or CI had been reprogrammed to provide no overlap of frequency information. In other words, the
CI could have provided only high-frequency information, or the HA could have
provided only low-frequency information.
6.3.2 Combining electric and acoustic stimulation in the same
ear
It is now the case that many CI recipients have some measurable unaided
hearing thresholds in both ears prior to implantation. The implantation
procedure usually results in a loss of some, if not all, of this residual hearing
(Boggess et al., 1989; Kiefer et al., 1998; Rizer et al., 1988). However, the
degree of hearing loss and the chance for hearing preservation after
implantation remain unclear. For example, Zappia et al. (1991) carried out an
histologic examination of the temporal bones of a multi-channel cochlear
implant patient. A large number of surviving hair cells were seen in the
implanted ear.
Of 45 Cochlear Nucleus implant recipients in Melbourne with at least one
measurable threshold pre-operatively, 49% were found to have at least one
measurable threshold post-operatively (Lie, 2003). The electrode array was
inserted fully into the cochlea for each of these subjects. The surgical procedure
was not modified in an attempt to preserve residual hearing. Hodges et al.
(1997) reported on a group of 40 CI subjects who had some measurable
hearing pre-operatively. Thirty-three of the subjects received Nucleus devices,
and seven received Clarion implants. Each array was inserted fully into the
cochlea. The surgical technique followed was in accordance with the
manufacturers’ recommendations and has been described by Balkany et al.
- 85 (1988). A measurable response for at least one frequency was found for 21 of
the 40 subjects post-operatively. On average, the group’s thresholds decreased
by 12 dB in the implanted ear. Sixteen (32%) out of a group of 50 cochlear
implant subjects were found to have some preserved residual hearing after
implantation (Shin et al., 1997). Seven of these patients showed no difference in
hearing thresholds between pre-and post-operative measures.
Skarzynski et al. (2002) reported success in preserving residual hearing by
using a modified surgerical technique, known as “soft surgery” (Lehnhart, 1993).
The method suggests using a minimal cochleostomy, reducing insertion depth,
and attempting to preserve perilymph. All patients received Med-El Combi
electrode arrays. Sixteen (62%) out of 26 patients retained their residual
hearing to within 5 dB of their pre-operative hearing thresholds by using this
surgerical approach. Five patients lost all hearing after the surgery. More recent
studies have focused on surgical techniques or modified electrode arrays in an
attempt to increase the success rate of hearing preservation in cochlear
implants. The results of these studies are summarized in Table 6.2. Each of
these studies is discussed in further detail below.
Authors
Year of
Publication
Gantz et al.
2005
Number of
subjects
implanted
24
von Ilberg
et al.
Skarzynski
et al.
Gstoettner
et al.
Kiefer et al.
1999
1
2003
1
2004
21
2005
13
James et
al.
2005
12
Electrode
array
Nucleus®
Hybrid™
Med El®
Combi 40+
Med El®
Combi 40+
Med El®
Combi 40+
Med El®
Combi 40+
Nucleus®
Contour
Advance
Hearing preserved to within:
10 dB
11-20 > 20 Total
dB
dB
loss
Average
1
1
loss for N
of 22 = 9
dB
1
1
13
1
8
3
5
4
3
2
5
Table 6.2 Success of hearing preservation in recently reported studies
An electrode array designed to stimulate the high-frequency basal region of the
cochlea while maintaining useful acoustic hearing in the low-frequency apical
region has been recently developed by Cochlear Limited. This has been
accomplished by employing a 10 mm, straight, intracochlear electrode array
2
- 86 – Chapter 6 Cochlear implantation and the benefits of residual hearing
attached to the Nucleus® 24 cochlear implant stimulator/receiver. This device is
called the Nucleus® Hybrid™ cochlear implant. The intracochlear electrode
array is shorter than a conventional electrode array (10 mm versus 24 mm), has
a smaller diameter, and has 6 active electrodes. The electrode array
incorporates a polyethyleneterephthalate (P.E.T.) mesh collar to prevent overinsertion, or further migration, into the cochlea beyond the point where the basal
turn curves into the ascending segment. Thus, the electrode array is placed
within the straight segment (first 10 mm) of the basal turn of the scala tympani.
The surgical technique used to insert the array is a modification of the technique
used to implant a standard array. Care is taken not to remove the inner ear
perilymph or interfere with the round window membrane (Gantz et al., 2003).
Initial results with the Hybrid device were reported for nine patients (Gantz et
al., 2003; Gantz et al., 2004). Initially the design of the electrode was 6 mm in
length with 6 active channels. Only 3 subjects received this electrode. The
remaining 6 subjects were implanted with an electrode of 10 mm in length,
which also had 6 active channels. The frequency allocation assigned to the
array was 2000 – 8000 Hz. Acoustic hearing was preserved for all nine patients
to within 10 – 15 dB. On average, subjects fitted with the 6 mm array obtained
an improvement in consonant recognition of 10% when the implant was worn
together with the hearing aid. Subjects fitted with the 10 mm array showed more
benefit, with consonant recognition improving by 40 percentage points on
average. For all subjects, there was a strong effect over time, with the greatest
improvements occurring 12 months post-operatively.
Preserving low-frequency acoustic hearing may have potential benefits for CI
users particularly for speech recognition in noise. Turner et al. (2004) compared
a group of 20 traditional implant recipients with 3 subjects who had been
implanted with a short-electrode array. Those 3 subjects all had some residual
acoustic hearing in the low frequencies post-operatively. The frequency range
assigned to the electrodes was 1062-7937 Hz for 2 subjects and 687-5187 Hz
for the third subject. The three subjects with the short-electrode array performed
better than the traditional implant group for speech recognition in a background
of competing talkers.
- 87 The 10 mm electrode array was used in a larger study in which 24 subjects
participated (Gantz et al., 2005). These subjects included the 9 CI recipients
whose results were reported in earlier studies (Gantz et al., 2003; Gantz et al.,
2004; Turner et al., 2004). Post-operatively, one subject had a total loss of
hearing and a second subject lost 30 dB of hearing. On average, the group lost
9 dB. At 6 months post-implantation, monosyllabic word understanding with the
CI together with the HA when compared with the binaural hearing aid condition
pre-operatively was significantly improved for 10 of the 11 subjects tested. Eight
subjects with the short-electrode array were compared with 20 traditional CI
subjects for testing in noise. The traditional CI subjects were selected so that
their average speech perception score in quiet was equal to that of the shortelectrode group. A signal-to-noise (SNR) ratio was determined at which each
subject obtained a 50% correct spondee score. Two-talker babble was selected
for the noise source. It was found that the short-electrode group performed
significantly better than the traditional CI subjects.
The results reported for the short-electrode array were promising, but could a
similar outcome be achieved with a longer standard array? The short-electrode
array may not have closely represented the normal tonotopic representation of
the cochlea. It’s likely that the array’s allocated frequency range was stimulating
at places associated with higher frequencies. This tonotopic shift may have
impacted on both subjects’ speech perception scores and the time taken to
adjust to the new electric stimulation.
One of the first reported cases of combined electrical and acoustic stimulation in
the same ear was described by von Ilberg et al. (1999). Cochlear implantation
was carried out on one patient with a steeply-sloping hearing loss. A Med-El
Combi 40+ electrode array was inserted 20 mm into the cochlea. Postoperatively, the patient showed a 20 dB drop in hearing thresholds for the
implanted ear. Only the 8 most-apical electrodes were able to be used for
stimulation. The remaining basal electrodes showed no auditory sensations at
maximum current levels. Speech perception testing was carried out at two
months post-operatively for three conditions: HA alone, CI alone, and CIHA in
the implanted ear. For the combined condition of CIHA a number of different
fittings were attempted. The number of electrodes was reduced from 8, to 6, to
- 88 – Chapter 6 Cochlear implantation and the benefits of residual hearing
5, to 4, and finally to 2. For each condition, apical electrodes were disabled until
only the 2 most-basal electrodes were stimulated. The acoustic frequencies
assigned to the electrodes were varied, ranging from 300 – 5500 Hz for the 8
electrode condition, 620 – 5500 Hz for the 6 electrode condition, 893 – 5500 Hz
for the 5 electrode condition, 1284 - 5500 Hz for the 4 electrode condition, and
1047 – 5500 Hz for the 2 electrode condition. There was no indication as to how
the given acoustic frequency ranges were selected.
For monosyllables, when all 8 electrodes were active, the patient scored 0% for
the hearing aid alone, 45% with the implant alone, and 50% for the implant
combined with the hearing aid. When the number of electrodes was reduced,
sentence scores for the CI alone condition reduced dramatically from 88% with
8 electrodes, to 18% with 6 electrodes to, 0% for 5 electrodes. For the CIHA
condition, sentence scores were 92% with 8 electrodes, 88% with 6 electrodes,
58% with 5 electrodes, 38% with 4 electrodes and 22% with 2 electrodes. It is
clear that this subject benefited overall from implantation, but the choice of
electrical stimulation was somewhat arbitrary. It is difficult to separate the
effects of reducing the number of electrodes from the acoustic frequency range
of stimulation. The patient may have obtained rather different speech results if
either the frequency range was varied but the number of active electrodes held
constant, or if the frequency range was held constant but the number of active
electrodes were varied.
Partial electrode insertion was reported to be successful with one subject who
had a steeply-sloping hearing loss (Skarzynski et al., 2003). Similarly to the
study above, a Med-El Combi 40+ electrode array was inserted 20 mm into the
cochlea. An audiogram measured post-operatively found a drop in hearing
thresholds of 15 dB at 500 and 1000 Hz, but no change in thresholds at
frequencies below 500 Hz. Before surgery the patient scored 20% on a test of
monosyllabic words in quiet. At 3 months post-operatively, the word score
obtained with the implant alone was 23%. However, the combination of electric
and acoustic stimulation resulted in a dramatic increase in speech perception,
with the patient scoring 90%. Recognition of speech in noise improved from 0%
before surgery to 65% after surgery with the implant together with acoustic
hearing. Based on these encouraging results, Skarzynski et al. (2003)
- 89 recommended that individuals with steeply-sloping losses be considered for a
cochlear implant if they fulfill certain criteria. These criteria included the
following: hearing thresholds of greater than 70 dB HL at frequencies above 500
Hz, and a monosyllabic word score of less than 55% in quiet for the best aided
condition.
Med El® Combi 40+ devices were implanted in 21 patients in a study reported
by Gstoettner et al. (2004). All patients had thresholds between 20 - 60 dB HL
at frequencies below 1000 Hz, and a severe-to-profound loss for frequencies
higher than 1000 Hz. The insertion depth of the electrode devices used in the
study ranged from 16 – 24 mm. The surgical technique involved performing a
mastoidectomy using a posterior tympanotomy approach, followed by an
atraumatic cochleostomy. In addition, steroids were applied both locally and
systemically. Preservation of low-frequency hearing was achieved in 18 of the
21 patients. A hearing loss of less than 10 dB was achieved for 13 patients.
Five patients showed a loss of greater than 10 dB. A total loss of residual
hearing in the implanted ear occurred in the remaining 3 patients. Hearing
thresholds remained stable over time post-operatively. Unfortunately, speech
perception results were reported for only one patient in the study. For this
patient, monosyllabic word scores measured at 3 months after surgery
improved from 35% pre-operatively, to 60% with the implant alone, and 90%
when wearing the implant together with a hearing aid in the implanted ear.
The optimum means of combining acoustic and electric stimulation has been
explored by Kiefer et al. (2005). Thirteen patients were implanted with the Med
El® Combi 40+ electrode array. An atraumatic surgical method was used as
described by Lehnhart (1993). The depth of the array was varied for each
patient using the frequency distribution map of Greenwood (1990) and the
patient’s audiogram. An insertion depth was calculated for each patient that
corresponded to the point at which the patient’s hearing thresholds were worse
than 65 dB HL. For the first 8 weeks post-operatively, the cochlear implant was
fitted to provide a frequency range of 300-5500 Hz. After this period, the implant
was programmed with 3 different programs. The first map provided a frequency
range of 300-5500 Hz, the second map was programmed to provide stimulation
at frequencies of 650 – 5500 Hz, and the third map was programmed to provide
- 90 – Chapter 6 Cochlear implantation and the benefits of residual hearing
stimulation at frequencies of 1000 – 5500 Hz. Subjects were encouraged to use
all 3 programs for 2-3 weeks, after which time speech perception testing was
carried out. In addition, subjects with residual hearing in the implanted ear were
fitted with ITE hearing aids.
Pre-operatively all subjects had thresholds of less than 60 dB HL at frequencies
below 500 Hz, and thresholds of greater than 60 dB HL at frequencies above
1000 Hz. Post-operatively, 11 of the 13 subjects retained some useful hearing.
For eight of these subjects, hearing was preserved on average to within 10 dB
of the original threshold. Three subjects lost between 11 – 20 dB, and 2
subjects lost all hearing in the implanted ear. Speech perception testing was
carried out at 12 months post-operatively with tests of monosyllabic words in
quiet and sentences in noise. The group improved on average from 7% with
bilateral hearing aids before surgery to 56% with the implant alone after surgery
and 62% in the CIHAi condition (cochlear implant together with hearing aid on
the ipsilateral side). All patients performed better with the CI alone compared to
the pre-operative HA score. Six patients performed better in the CIHAi condition
when compared with the CI alone score. Four patients showed no difference in
scores between the CI alone and CIHAi conditions. One patient obtained a
lower score for the CIHAi condition when compared to the CI alone condition.
For testing in noise, 7 out of 12 patients performed best in the CIHAi condition.
Three patients showed no difference between the CIHAi and CI alone
conditions.
The large majority of patients obtained the best performance with the map
which provided the broadest frequency range of 300-5500 Hz. One patient
preferred the map which provided a frequency range of 650 – 5500 Hz. This
result may have been confounded by the amount of time patients were exposed
to the different maps. Patients received 10 weeks’ experience with the first map,
compared to just 2-3 weeks with the other two maps. Patients thus had a longer
time period to adjust to the sound quality of map 1.
Finally, several clinics around Europe are currently participating in a multicentre trial investigating the conservation of residual hearing in subjects who
are borderline candidates for cochlear implantation (James et al., 2005).
Inclusion criteria are described in terms of speech understanding rather than
- 91 hearing thresholds. To be included in the study all patients must have a
minimum of 10% word understanding in the ear to be implanted. The maximum
word score considered for implantation was not given. It is hoped that 100
subjects will be eventually recruited for the study. Twelve patients were
implanted with the Nucleus Contour Advance perimodiolar electrode array. A
soft-surgery technique was followed which consisted of a relatively small
cochleostomy of 1-1.2 mm located anterior and inferior to the round window,
and a reduced insertion depth of 17 mm. Of the 12 subjects, 2 had a total loss
of hearing in the implanted ear as a result of the surgery. For the remaining 10
subjects, the average increase in hearing thresholds was 23-33 dB for the
frequency range 125-1000 Hz.
Subjects were divided into two groups depending on their post-operative
hearing thresholds. If hearing thresholds were better than or equal to 80 dB HL
for the frequencies 125, 250 and 500 Hz, subjects were classified as being
“electric-acoustic” users. These subjects were fitted with an ITE hearing aid in
the implanted ear. Two maps were created: map A which was a conventional
map with the full-frequency range, and map C in which high frequencies were
assigned in the processor to begin from the highest frequency at which
subjects’ acoustic hearing ended. Users were given one of these two maps for 4
weeks post-operatively, and then switched to the alternative map for a further 4
weeks. At 2 months subjects were allowed to switch between the two maps.
Speech perception results measured at 3 months post-operatively were only
reported for the map in which the subject performed best. Subjects who had
thresholds worse than 80 dB HL were treated as traditional CI patients.
Therefore, all these subjects were given map A (i.e. the conventional map).
Speech recognition results have been reported for 6 of the 12 patients. Three of
these patients were traditional CI patients, whereas the remaining three were
electric-acoustic users. Results in quiet found that 4 of the 6 subjects performed
better with the CI alone when compared with their pre-operative scores. One
subject (P8) in the electric-acoustic user group, who obtained a word score of
30% with HAs pre-operatively, obtained chance scores when wearing the CI
alone. All three of the electric-acoustic users performed best when wearing the
implant together with the ipsilateral hearing aid (CIHAi). Subject P8 showed no
- 92 – Chapter 6 Cochlear implantation and the benefits of residual hearing
significant differences in scores between the pre-operative HA condition and
post-operative CIHAi condition. In noise, 5 of the subjects showed
improvements from their preoperative scores when wearing the CI alone. Two
users in the electric-acoustic group performed best when wearing the implant
together with the ipsilateral hearing aid. When compared to pre-operative HA
scores, subject P8 in the electric-acoustic group performed worse in noise postoperatively for both the CI alone and CIHAi conditions.
Unfortunately, for the James et al. (2005) study, the reader is not provided with
any detail as to what map the electric-acoustic users were using for the speech
recognition measures reported. This study and much of the previously reported
work on this topic has been focused on hearing preservation and surgery
techniques. It answers the question that hearing preservation is possible with a
standard electrode array although the risk of some hearing loss, if not a total
loss, remains. However, mapping techniques need to be more formally
investigated. So far, an assumption has been made that the frequency range
allocated in the processor should begin at the highest frequency at which
acoustic hearing is no longer seen to be useful. This frequency has been
selected from visual inspection of the audiogram or Greenwood’s calculation. It
should be determined if this mapping would result in the best outcome for
patients, and if so, a more formal technique of selecting the input-frequency
range of the processor should be developed.
- 93 -
At the time of writing, if all these studies are grouped
together, a total of 72 implantation cases have been
reported in which there has been an attempt to preserve
residual hearing. Of these 72, a total of 8 (11%) patients
lost all hearing as a result of the surgery. The large
majority of patients showed a drop in hearing thresholds
post-operatively. Speech perception results have only
been reported for 32 of the 72 patients. Of these 32, a
total of 31 (97%) showed improvements in speech
understanding when wearing the implant together with a
hearing aid. Positive results have been reported for both a
short-electrode array, as well as partial insertion with a
standard-electrode array. In the current study, discussed
in Chapter 9, a group of listeners with steeply-sloping
losses were implanted with the Nucleus Freedom device.
The surgical method used for the subjects attempted to
preserve acoustic hearing. Two fitting strategies were
tested. For the first strategy, the implant was programmed
to provide the listener with high-frequency information
above the frequency at which the acoustic hearing was no
longer
considered
useful.
The
acoustic
frequency
corresponding to the most-apical electrode in terms of the
pitch perceived was determined with psychophysical
measures described in Chapter 9. This frequency was
used to create the first fitting program for each of the
subjects which attempted to mimic the ear’s frequencyplace relationship. The second strategy provided electrical
stimulation for the full input frequency range.
- 94 – Chapter 6 Cochlear implantation and the benefits of residual hearing
- 95 -
Chapter 7
Experiment 1
Benefits of audibility for listeners with severe
high-frequency hearing loss
7.1
Introduction
Previous chapters have summarized that individuals with sloping losses often
do not perceive important high-frequency information due to the nature and
configuration of their hearing loss. This thesis is primarily concerned with
investigating methods of providing this information. The simplest means of
achieving this was to amplify the high-frequency components of the signal. The
amplified signal was then presented to the listener. The first experiment of the
current study examined what effects high-frequency amplification had on
speech perception in a group of hearing-impaired listeners. Ten subjects with
moderate-to-profound high-frequency losses were selected to participate.
Further information regarding the subjects is provided in section 7.2. The TEN
test was carried out with all subjects. The methods and procedures followed for
Experiment 1 is provided in section 7.3. Speech perception results under
various low-pass filter conditions are shown in section 7.4. The results of this
experiment have been published (Simpson et al., 2005a). A copy of the
publication can be seen in Appendix B.
7.2
Subjects
Ten hearing-impaired adults (S01- S10) participated in the study. They were
recruited via the Cooperative Research Centre (CRC) for Cochlear Implant and
Hearing Aid Innovation. All subjects were participants in other research projects
at the time of testing.
7.2.1 Criteria for recruitment
The criteria considered when approaching potential participants were developed
with the aim of recruiting subjects with potential dead regions. They were as
follows:
- 96 – Chapter 7 Benefits of audibility for listeners with severe high-frequency hearing loss
1. Post–linguistic, sensorineural hearing impairment. No air-bone gap or
conductive component was found in any subject at the time of testing. All
subjects had acquired language well before the onset of their hearing
impairment.
2. Moderate-to-profound high-frequency hearing impairment. All subjects
had sloping high-frequency audiograms with thresholds generally worse at
higher frequencies than at lower frequencies. Only one ear was used as the test
ear for each subject. If the subject presented with an asymmetrical hearing
impairment, the ear with the better hearing thresholds was selected to be the
test ear.
3. Previous hearing-aid experience. All subjects were experienced hearingaid users. At the time of the study, all subjects were participating in a separate
hearing-aid trial. As a result, all subjects were wearing the same behind-the-ear
(BTE) hearing instrument, namely the Phonak Supero 412. They had been
wearing this instrument for a period of several months prior to the
commencement of Experiment 1.
Figure 7.1 shows each subject’s air-conduction thresholds for the test ear.
Further information regarding each subject can be found in Table 7.1.
7.2.2 Ethics approval
This study was approved by the Human Research and Ethics Committee at the
Royal Victorian Eye and Ear Hospital (Project No 94/243 H). The test procedure
and aims of the study were discussed with each subject. Written information
regarding the study was given to each subject (Appendix A). Informed consent
was obtained from each subject. Each subject was reimbursed for their travel
costs by the Cooperative Research Centre (CRC) for Cochlear Implant and
Hearing Aid Innovation.
- 97 -
0
S01
S02
S03
S04
S05
S06
S07
S08
S09
S10
Hearing level (dB HL)
20
40
60
80
100
120
0.25
0.5
1
1.5
2
3
4
8
Frequency (kHz)
Figure 7.1 Air-conduction thresholds in dB HL for all subjects. Arrows indicate no
response at the audiometer’s maximum output.
Ear
Subject tested
Age
Sex
Etiology
S01
R
82
M
Industrial Noise Exposure
S02
L
63
M
Industrial Noise Exposure
S03
R
59
F
Otosclerosis
S04
R
60
F
Viral Infection
S05
R
80
M
Industrial Noise Exposure
S06
L
69
M
Industrial Noise Exposure
S07
R
74
M
Unknown
S08
L
74
M
Industrial Noise Exposure
S09
R
69
M
Industrial Noise Exposure
S10
R
69
M
Industrial Noise Exposure
Table 7.1 Relevant information regarding the subjects who participated in Experiment 1.
- 98 – Chapter 7 Benefits of audibility for listeners with severe high-frequency hearing loss
7.3
Methods and procedures
7.3.1
Introduction
The focus of section 7.3 is to describe the tools and methods used when
collecting the data for Experiment 1. The section is divided into four. Section
7.3.2 describes the procedure followed when obtaining general audiological
data about each subject. Section 7.3.3 is concerned with the method followed
when measuring dead regions in the cochlea. The method used to determine
audibility of the consonant stimuli for each subject is the focus of section 7.3.4.
Finally, section 7.3.5 describes the selected speech recognition test.
7.3.2
Measurement of hearing thresholds
The hearing-aid usage and medical history of each subject was documented
during the first test session before the measurement of hearing thresholds.
Hearing thresholds were determined as a means for selecting subjects with
possible dead regions.
Stimuli
The stimuli consisted of pure tones with duration of approximately one second.
The Madsen Aurical audiometer was used to generate the stimuli.
Method
Subjects were seated in a medium sized sound-proof booth. Subjects made use
of a response button to indicate when they heard a stimulus. Hearing threshold
levels in dB HL were measured conventionally under Telephonics TDH39
headphones. The tone levels were adjusted manually, with the procedure
proposed by Carhart and Jerger (1959). The following frequencies were tested:
250, 500, 1000, 1500, 2000, 3000, 4000, and 8000 Hz. A threshold at 750 Hz
was also measured if the difference between the thresholds measured at 500
and 1000 Hz was greater than or equal to 20 dB.
7.3.3 Measurement of dead regions
The aim of the dead regions test is to diagnose regions in the cochlea where
there are no surviving inner hair cells and/or neurons (Moore et al., 2000a). It is
also referred to as the TEN (threshold equalizing noise) test. The TEN test was
chosen for this study as it is currently, apart from speech recognition measures,
the most recognized clinically available test to diagnose dead regions. As
- 99 discussed in Chapter 3, it is based upon the detection of sinusoids in the
presence of a broadband noise, designed to produce almost equal masked
thresholds (in dB SPL) over a wide frequency range.
Stimuli
All stimuli used in testing were obtained from the CD “Diagnosis of Dead
Regions” developed by Moore et al. (2000b). The CD recording contained two
channels. The first channel had several different test signals consisting of
digitally generated pure tones at the following frequencies: 250, 500, 1000,
1500, 2000, 3000, 4000, 5000, 6000, 8000, and 10 000 Hz. Each test frequency
was assigned a unique track number on the CD.
The second channel on the CD was a spectrally-shaped broadband noise
termed “threshold equalizing noise” (TEN). The TEN noise is intended to
provide equal masked thresholds in dB SPL by producing a constant amount of
excitation at each CF in the mid-frequency range of 500 to 5000 Hz. Moore
(2001) developed the required spectral characteristics of TEN by means of the
following formula:
Ps = N o .K .ERB,
Equation 2
where the power of the signal at threshold is Ps, the noise power spectral
density is N o , K is the signal-to-noise ratio at the output of the auditory filter
required for threshold, and ERB is the equivalent rectangular bandwidth of the
auditory filter. The TEN was spectrally shaped so that N o .K .ERB was constant
over the frequency range of 125 – 15000 Hz. Values of K and ERB were taken
from Moore & Glasberg (1997). The noise level is specified in terms of the level
in a one-ERB wide band around 1000 Hz (i.e. the level in the frequency range
935 to 1065 Hz). An example of the noise spectrum is shown in Figure 7.2 for a
level of 70 dB/ERB. The masked threshold is approximately equal to the
nominal level of the noise specified in dB/ERB. For example, a level per ERB of
50 dB will approximate a masked threshold of 50 dB SPL.
- 100 – Chapter 7 Benefits of audibility for listeners with severe high-frequency hearing loss
Method
Moore et al. (2000b) recommended a method for carrying out the TEN test on a
conventional audiometer. This method was modified to enable testing to be
carried out on the Madsen Aurical audiometer. The CD was played through a
Pioneer Multiplay Compact Disc Player (PD-M423). The levels of the stimuli
were adjusted on the Audiometer Speech Module Screen. Testing was carried
out under Telephonics TDH39 headphones. Subjects made use of a response
button to indicate when they heard a test stimulus. The test instructions advise
the clinician to set both VU meters on the audiometer to read –6 dB prior to
commencing. The Aurical software limits the VU meters to a maximum of –5 dB.
This 1 dB difference was taken into account when evaluating results.
Figure 7.2 Spectrum of the TEN for 70 dB/ERB (Figure reproduced from Moore, 2001).
7.3.3.1.1 Measurement of absolute thresholds
Subjects were given the following instructions:
”You will be hearing a series of sounds. These sounds will vary in pitch and
loudness. Press the response button immediately after you hear one of these
sounds, even if the sound is very soft. Press the button even if you think you
heard a sound.”
- 101 The standard audiometric method recommended for measuring pure-tone
thresholds (Carhart et al., 1959) was used to determine absolute thresholds in
dB SPL. Absolute thresholds were determined in the test ear at the following
frequencies: 250, 500, 1000, 1500, 2000, 3000, 4000, and 5000 Hz. The order
of presenting test frequencies was randomized across subjects and test
sessions.
7.3.3.1.2 Measurement of masked thresholds
A masking noise level that was higher than the lowest absolute threshold
measured at any frequency was selected as the initial level. The subject was
given the following instructions:
“You will be hearing a continuous wind-like noise. Try to ignore this noise. A
second series of short sounds will be played as well. These sounds will vary in
loudness and pitch. Press the button only when you hear one of these sounds,
even if the sound is very faint. If you think you have heard a sound, you may
also press the button.”
Masked thresholds were determined next at each test frequency. The same
method was used as when measuring absolute threshold with one exception.
When nearing threshold, 1 dB incremental steps were used instead of 5 dB
steps to obtain a more precise measurement.
Wherever possible, the noise level was increased and the masked thresholds
were measured again to obtain measurements at more than one level/ERB. A
high masking level was required to mask high frequencies as all subjects had
severe high-frequency hearing losses. For this reason, TEN levels of 81 and 91
dB/ERB were selected when testing most subjects.
The order of presenting test frequencies was randomized across subjects and
test sessions. The entire procedure of measuring both absolute and masked
thresholds was repeated at the following test session to obtain two sets of
threshold data. These two sets of data were averaged for later analysis.
7.3.4 Measurement of aided thresholds
Aided thresholds were selected as a measure of audibility of the speech stimuli
for each subject. To maximize audibility of the speech stimuli, direct audio input
was selected as the means of presenting stimuli to each subject. The use of
- 102 – Chapter 7 Benefits of audibility for listeners with severe high-frequency hearing loss
direct audio input ensured that no acoustic feedback could occur with the
hearing aid, and therefore enabled the gain to be set to higher than normal
values. Before describing how aided thresholds were measured, it is necessary
to first discuss the method followed for enabling testing to be carried out via
direct audio input.
All further testing was performed with each subject wearing a test hearing
instrument, namely the Phonak Supero 412. These hearing aids were specified
to have a maximum output and a maximum gain of approximately 140 dB SPL
and 80 dB, respectively (measured in an ear simulator), and were designed to
be most suitable for people with hearing threshold levels that exceed 50 dB HL
at all frequencies. The gain was separately adjustable in five partiallyoverlapping frequency bands.
The Phonak AS3/AS4 Audioshoe was selected to present the stimuli via direct
audio input. The audioshoe disconnected the microphone and delivered electric
signals directly to the input of the aid’s amplifier. The subject’s aided thresholds
were measured when direct audio input had been enabled. The procedure
followed when calibrating the audioshoe will be described first in section 7.3.4.1.
The procedure followed when measuring aided thresholds through the
audioshoe is described in section 7.3.4.2.
7.3.4.1 Audio shoe calibration
The audioshoe was calibrated to be compatible with two software programs: (a)
MACarena Speech Tests 2 and (b) Presentation 3. MACarena was used to
present each subject with consonant stimuli. Presentation was used to measure
aided thresholds.
2
MACarena Speech Tests is a software program designed to present speech stimuli and record
subject responses. It was developed by Dr. Waikong Lai from the University of Zurich. Copyright
USZ_ORLLEA (May 2003).
3
Presentation is a stimulus delivery and experiment control software system. It was modified for
in-house use by Gerhard Hannemann. The current experiment was performed using
Presentation® software Version 0.70. The program can be downloaded from
www.neurobs.com.
- 103 7.3.4.1.1 Calibrating the audioshoe for use with the MACarena software
program
The Supero hearing aid was connected to a compatible fitting software
program. The program selected for Experiment 1 was the Phonak Fitting
Guideline (PFG), Version 8.0. A fitting suggestion, based on each subject’s
audiogram, was automatically generated by the PFG using NAL-RP (Byrne et
al., 1986) fitting guidelines. Some adjustments were made to the program to
turn off any features in the hearing instrument which may have affected the
calibration. These adjustments included selecting a linear-processing strategy,
turning off the noise canceller, and deactivating the feedback canceller.
All adjustments to the hearing instrument were saved before exiting the fitting
program. The Supero was switched on and connected to a Bruel & Kjaer Ear
Simulator Type 4157 prior to running the calibration measurement. The ear
simulator was connected to a Bruel & Kjaer Pulse Sound Analyser Type 3560C.
The hearing instrument was placed at 1 m from a Tannoy Reveal loudspeaker.
The microphone of the hearing aid was placed at approximately the same
height as a subject’s ear would be. The MACarena program was run. The
broadband calibration noise of MACarena was played through the loudspeaker.
This noise was based on the international long-term average speech spectrum
(LTASS) according to Byrne et al. (1994). A Bruel & Kjaer Type 2239 soundlevel meter set to frequency-weighting A, fast (level averaging with a time
constant of 125 ms), and maximum (displaying the maximum value recorded
within the preceding second) was used to measure the level of the calibration
noise. A level of 60 dBA, which corresponds to a level between conversational
and raised-to-loud speech (Skinner et al., 1997), was recorded at 1m from the
loudspeaker at the time of testing.
There are four ways in which the level of this noise could be varied: (1) in the
Windows software mixer, using the Master Volume control; (2) in the Windows
software mixer, using the Wave Volume control; (3) in the MACarena program;
and (4) using the amplifier’s volume control. To minimize the chance of
accidentally varying the levels, all but one of these controls was set to a
predetermined value. In this study, the MACarena program was used to set the
test levels while keeping all other controls at a constant value. The MACarena
- 104 – Chapter 7 Benefits of audibility for listeners with severe high-frequency hearing loss
level slider showed a value of -16 dB at the time of testing. A Bruel & Kjaer
Pulse Sound Analyser Software Program Type 7700 was run to measure the
output of the Supero via the ear simulator while the calibration noise was
playing. Figure 7.3 shows the resulting output curve.
The hearing instrument was then attached to the audioshoe. The audioshoe
cable was fed directly from the soundcard of the computer. The MACarena
program was run. The broadband calibration noise described previously was
played while measuring a second output curve. The MACarena level slider was
adjusted until the output curve of the audioshoe closely matched the output
curve measured in the sound field. The sum of the squared errors was
minimized to determine the setting which most closely approximated the
soundfield output curve.
These new settings were noted. The MACarena level slider value at the time of
testing was -8 dB. The volume controls of the CD player and computer sound
card were kept constant. Figure 7.3 shows the recorded output curves.
dBSPL (ear sim) re 20 uPa
100
90
80
70
60
Soundfield
50
40
Audioshoe
30
20
10
0
100
1000
10000
Frequency (Hz)
Figure 7.3 Frequency response of the Supero 412 hearing instrument in response to the
MACarena broadband calibration noise at a level of 60 dBA. The Supero’s response was
recorded via an ear simulator under two test conditions: (a) in the soundfield (solid line),
and (b) via direct audio input (dashed line).
The audioshoe could now be used with the MACarena program. A level
equivalent to 60 dBA in the soundfield was achieved using the audioshoe by
- 105 changing the level slider value of the MACarena calibration noise accordingly:
i.e. a value of –8 dB through the audioshoe was equivalent to a value of –16 dB
through the soundfield.
7.3.4.1.2 Calibrating the audioshoe for use with the Presentation software
program
A similar procedure was followed as described above in section 7.3.4.1.1. The
Supero was switched on and connected to the ear simulator. The hearing
instrument was placed at approximately 1 m from the speaker in the soundfield.
The MACarena software program was run. The calibration noise from
MACarena was played through the loudspeaker. The characteristics of this
noise are described in the section above. The level of the noise was adjusted
until it measured 60 dBA at 1m from the loudspeaker. A sound-level meter with
the following settings was used to perform the measurement: frequencyweighting A, fast, and maximum. The MACarena level slider was set to a value
of -16 dB at the time of testing. The calibration noise was then turned off.
The Presentation Software program generated one-third octave narrowband
noises to measure aided thresholds. The levels of these narrowband noises are
based on the LTASS according to Byrne et al. (1994). The following
narrowband noises were selected: 250, 500, 1000, 1600, 2000, 3150, and 4000
Hz. The narrowband noise duration of 0.5 sec was too brief to obtain a sound
level measurement. Instead, the CoolEdit Pro (Version 2.0, Syntrillium Software
Corporation) software program was used to play the stimuli while calibrating
their levels. The signal was selected and looped to play continuously through
the loudspeaker. The dB SPL level of the narrowband noise was measured with
a sound level meter at 1m from the loudspeaker. The sound level meter was set
to the following: fast, linear (flat-frequency response), and one-third octave
band-pass filters. The output level from the hearing aid was also recorded. This
measurement was carried out with each of the narrowband noises. Table 7.2
shows the levels that were recorded.
The hearing aid was attached to the audioshoe which was connected directly to
the computer sound card. The MACarena level slider was turned to the point
previously calculated to equal 60 dBA, i.e. –8 dB. The Pulse Software program
was run to measure the output of the hearing aid as each narrowband noise
- 106 – Chapter 7 Benefits of audibility for listeners with severe high-frequency hearing loss
was played through CoolEdit. These levels were recorded, and listed in Table
7.2.
Some differences between the soundfield and the hearing-aid output levels
were noted (see far right column of Table 7.2). These differences were included
as correction factors when calculating aided thresholds for specific frequencies.
Frequency (Hz)
250
500
1000
1600
2000
3150
4000
Hearing-aid output measured
with the audioshoe input
(dB SPL)
76.9
88.4
80.5
86.9
90.3
91.2
89
Hearing-aid output
measured in soundfield
(dB SPL)
76.8
85.4
84.2
86.6
88.5
92.3
92
Difference
(dB)
- 0.1
-3
3.7
- 0.3
- 1.8
1.1
3
Table 7.2 Recorded output levels of the Supero 412 via an ear simulator, for input levels
corresponding to the LTASS (Byrne et al., 1994).
7.3.4.2 Measuring aided thresholds through the audioshoe
7.3.4.2.1 Supero gain adjustment
As mentioned, Experiment 1 aimed to maximize the audibility of the speech
stimuli for each subject without causing discomfort. To achieve this, aided
thresholds needed to fall below the minimum speech levels across the
frequency range. For the purposes of this study, these levels were taken to be
18 dB below each one-third octave band level of the LTASS according to Byrne
et al (1994).
The amount of gain necessary for aided thresholds to fall below the speech
minima in each one-third octave band was determined for each subject using
the following formula:
G ( f ) = ACT ( f ) + C ( f ) − S ( f ) ,
Equation 3
where G was the gain in dB for a 2cc coupler, ACT was the air-conduction
threshold in dB HL, C was the conversion factor from dB HL to dB SPL for a 2cc
- 107 coupler (Bentler et al., 1989), S was the minimum one-third octave band level
for speech at an overall level of 60 dB SPL (Byrne et al., 1994), and f was the
frequency. Table 7.3 shows the values used to determine the gain settings for
each subject.
Frequency (Hz)
250
500
1000
2000
4000
Conversion factor HL to 2cc SPL
(dB)
12.2
7.5
4.3
7.9
1.2
Minimum LTASS (60dB SPL)
35
37
28
20
23
Table 7.3 Table used to determine the amount of gain necessary to achieve aided
thresholds 18 dB below each one-third octave band level of the LTASS (Byrne et al.,
1994).
Before measuring aided thresholds, the Supero was connected to the Phonak
Fitting Guideline (PFG) fitting software and the gain adjusted to be as close as
possible to the gain settings calculated by means of the formula above. In the
hearing aid, a linear-amplification scheme was selected, and the noise canceller
and feedback canceller were deactivated. In some cases, particularly at very
high frequencies, it was not possible to achieve a threshold below the speech
spectrum. In these cases, the gain was set to the maximum value allowed by
the fitting software. All subsequent testing was performed with each subject
wearing this hearing aid fitted in accordance with their individually determined
program.
Stimuli
One-third octave narrowband noises with duration of 0.5 sec were selected for
measuring aided thresholds. The noises were ramped in level at each end with
linear ramps of duration 30 ms. The level of each narrowband noise was set to
equal the one-third octave level at the corresponding frequency of the LTASS,
according to Byrne et al. (1994). The frequencies selected for this test were:
250, 500, 1000, 1600, 2000, 3150, and 4000 Hz.
Method
The Presentation Program was used to present the stimuli to the subject.
Subjects were tested individually. The Supero was adjusted as described in
- 108 – Chapter 7 Benefits of audibility for listeners with severe high-frequency hearing loss
section 7.3.4.2.1 and attached to the audioshoe. The audioshoe cable was
connected to the computer soundcard. The Supero was connected to the
subject’s own ear-mold and placed in the test ear. Prior to testing, the vent on
each subject’s mold was blocked by inserting a closed plug to minimize sound
leakage. The MACarena level slider was set to the appropriate level of –8 dB
prior to testing. Subjects were seated at a desk facing a computer monitor. They
were given the following instructions:
“You will see a series of four red boxes numbered 1-4 displayed on the
computer screen in front of you. Each box lights up in order. Click on the box
you think corresponds to the sound being presented using the computer mouse.
It is possible that you may not hear a sound presented with any of the four
boxes. You must guess in this case.”
A narrowband noise was selected in the Presentation program. A series of four
red boxes were displayed on the computer screen, numbered 1-4. The test was
presented using a four interval forced choice (4IFC) method. Each box was lit
up in order (1 to 4). The subject was required to pick which of the 4 boxes
corresponded to the sound being presented.
Each subject’s aided threshold for each frequency was determined using an
adaptive procedure. The initial level (described above under Stimuli) was
presented twice by Presentation. There was a 4 dB reduction of the level once
two correct responses were recorded in a row. The reduction in SPL continued
in this way until a point was reached where the subject made an incorrect
response. Once one incorrect response was made, the SPL was increased by 4
dB. This point was automatically recorded as the first turning point. Once two
turning points were obtained, the incremental change in dB was reduced to 2
dB. A further 6 turning points were obtained using steps of 2 dB. Presentation
calculated the average reduction in SPL over the last 6 turning points. Results
were automatically recorded in the computer.
To determine what the actual aided threshold was, the average reduction in
SPL was subtracted from the initial starting level (in dB SPL). The subject’s
aided threshold when measured through the audioshoe was calculated by
means of the following formula:
- 109 -
AT ( f ) = S ( f ) − d ( f ) + r ( f )
Equation 4
where AT was the aided threshold in dB SPL, S was the minimum one-third
octave band level for speech at an overall level of 60 dB SPL (Byrne et al.,
1994), d was the difference between the output levels measured in the
audioshoe and in the soundfield, r was the average result recorded in
Presentation, and f was the frequency.
This procedure was repeated for each frequency. Initially one measurement
was obtained at each frequency. The turning points for each frequency were
analysed statistically using the software program Sigma Stat (Version 1.0,
Jandel Corporation). The measurement was repeated at a particular frequency
if the standard error of all 8 turning points was greater than or equal to 2.0 dB.
7.3.5 Speech recognition testing
Stimuli
The test materials comprised vowel-consonant-vowel (VCV) nonsense syllables
recorded by a male and a female speaker. The 16 consonants selected for the
test were: /p/, /t/, /k/, /b/, /d/, /g/, /m/, /n/, /s/, /S/, /z/, /f/, /v/, /tS /, /j/, and /T/. The
initial vowel was always the same as the final vowel.
The male-speaker stimuli were selected from recordings on the NAL CRC CD,
disc 1 – “Speech and Noise for Hearing Aid Evaluation” (National Acoustic
Laboratories., 2000). They were embedded in the vowels /a/ and /i/, and spoken
with a typical Australian accent. The female-speaker stimuli were obtained from
University College London (Markham et al., 2002).
These stimuli were low-pass filtered at four cut-off frequencies: 1400, 2000,
2800, and 5600 Hz. A 10th order Butterworth filter was selected with a slope of 60 dB/octave. The levels of the medial consonant tokens were approximately
equalized before filtering. The average levels were approximately 60 dBA. The
total number of stimuli was 256 (i.e., 16 /aCa/ consonants + 16 /iCi/ consonants
× 4 low-pass filter conditions × 2 speakers). Each stimulus was repeated four
times making a total of 1024 tokens.
- 110 – Chapter 7 Benefits of audibility for listeners with severe high-frequency hearing loss
Method
A closed-set test was run, in which 16 buttons appeared on the MACarena
screen. These buttons were of the following format: vCv where the vowel was
either /i/ or /a/, and C represents each consonant. The positions of these
buttons were fixed for each subject to allow the subject to become familiar with
the response options. Figure 7.4 shows how the MACarena screen appeared to
each subject during testing.
The stimuli were divided into 16 subtests consisting of the 4 low-pass filter
conditions × 2 speakers × 2 vowels. Each stimulus was repeated twice in each
subtest. Each subtest was repeated twice to make a total of 4 repetitions for
each token. MACarena presented the stimuli in a random order in each subtest.
The stimuli were divided into 4 blocks: female speaker with /a/ vowel, female
speaker with /i/ vowel, male speaker with /a/ vowel, and male speaker with /i/
vowel. The 4 blocks each contained 4 subtests which were the 4 low-pass filter
conditions. One block was presented to each subject at each test session. The
entire test was completed over 4 test sessions. The order in which these blocks
were presented was randomized across subjects and test sessions. The lowpass filter conditions were presented in the following order within each of the
four blocks: 5600, 2800, 2000, 1400, 1400, 2000, 2800, and 5600 Hz.
Reversing the order of the filter conditions helped to reduce any learning effects
during the testing procedure.
The consonant stimuli were discussed with each subject prior to testing. The
tester showed the subject each button and verbalized which sound
corresponded to the button on the screen. Once the tester was confident the
subject was comfortable with the task, testing began. Subjects were seated at a
desk facing a computer touch-screen monitor. The stimuli from MACarena were
fed directly into the audioshoe connected to the subject’s hearing aid. Any vent
in the subject’s mould was blocked prior to presenting the stimuli.
The subject was given the following instructions:
“You will be hearing a series of consonants. Each consonant will be
immediately preceded and followed by the vowel /a/ or /i/. Press the box you
think corresponds to the consonant being presented. You are allowed to guess
- 111 as some of the consonants may sound soft or distorted. It is possible that one
consonant may be repeated several times. You may not hear all of the
consonants on the screen in front of you.”
Figure 7.4 Screen shot of the MACarena test format.
At the start of each test session, each subject performed the easiest subtest
(i.e. male speaker, cut-off of 5600 Hz, using the vowel /a/) to gain practice and
confidence in the procedure. The results were recorded as usual, but
disregarded in the final analysis of the data.
7.4
Results
Figure 7.5 shows the TEN test results for the group of subjects tested in
Experiment 1. Absolute thresholds are shown for each subject in dB SPL across
the frequency range. Measured masked thresholds in dB SPL at 250, 500,
1000, 1500, 2000, 3000, 4000, and 5000 Hz are shown for each subject at two
masking levels/ERB. Each graph has two shaded areas that are 10 dB in width.
- 112 – Chapter 7 Benefits of audibility for listeners with severe high-frequency hearing loss
0
20
40
60
80
100
120
0
S01
S02
S03
S04
20
40
60
80
100
120
0
Level (dB SPL)
20
40
60
80
100
S06
S05
120
0
20
40
60
80
100
S07
S08
S09
S10
120
0
20
40
60
80
100
120
0.25
0.5
1 1.5 2
3 45
0.25
0.5
1 1.5 2
3 4 5
Frequency (kHz)
Figure 7.5 TEN test results for S01-S10. Absolute thresholds in dB SPL for each subject
are represented by filled triangles across a frequency range of 250-5000 Hz. Masked
thresholds at two noise levels (filled and unfilled circles) are shown for each individual
subject in dB SPL across the same frequency range. Error bars on each symbol
represent plus and minus one standard deviation from the mean. The two shaded areas
on each graph are 10 dB in width. They represent graphically the point outside of which a
masked threshold is considered to indicate a dead region according to TEN test
principles. Solid black horizontal lines represent measured dead regions for S01, S02,
and S06.
- 113 These areas represent graphically the point outside of which a masked
threshold was considered to indicate a dead region. For example, a measured
masked threshold of 94 dB SPL measured with a noise level/ERB of 81 dB/ERB
fell outside of the first shaded area and was therefore considered to indicate a
dead region. A measured masked threshold of 94 dB SPL measured with a
noise level/ERB of 91 dB/ERB fell within the second shaded area and was not
considered a dead region.
Three subjects (S01, S02, S06) showed dead regions according to the TEN test
criteria. For these subjects, masked thresholds were elevated above both
measured absolute thresholds and the masking noise level. S06 showed
elevated masked thresholds at 2000, 3000, and 4000 Hz. S02 showed an
elevated threshold measurement at 4000 Hz only. Measured thresholds for
these two subjects (S02, S06) showed a similar pattern of elevation at both
noise levels of 81 and 91 dB/ERB.
S01’s masked threshold at 1500 Hz was elevated when the noise level was 91
dB/ERB. The threshold was not elevated at the lower masking noise level of 81
dB/ERB. No other indications of dead regions were found for this subject at
other tested frequencies.
Dead regions were not found at any frequency with the remaining 7 subjects.
One subject (S08) showed approximately equal masked thresholds across the
frequency range. According to TEN test principles it can be assumed that this
subject had no dead regions. For subjects S04, S05, and S07, the TEN noise
was not sufficiently intense to produce 10 dB or more of masking above 3000
Hz.
For subjects S03, S09, and S10, the TEN noise was not sufficiently intense to
produce 10 dB or more of masking above 4000 Hz. For example, S07 had high
absolute thresholds at 4000 and 5000 Hz. Even the highest TEN level of 91
dB/ERB was not sufficient to mask these frequencies. It was impossible to
determine with the TEN test whether or not these subjects had dead regions at
these high frequencies.
To summarize, 3 subjects (S01, S02, S06) had dead regions according to the
TEN test. It was not possible to determine whether 6 subjects (S03, S04, S05,
- 114 – Chapter 7 Benefits of audibility for listeners with severe high-frequency hearing loss
S07, S09, S10) had dead regions because of limitations of the TEN test. One
subject (S08) appeared to have no dead regions.
Figure 7.6 shows each subject’s aided thresholds in dB SPL across the
frequency range as measured through the audioshoe. The shaded area
represents a “speech banana”. Values are adapted from the LTASS (Byrne et
al., 1994) for speech measured at an overall average level of 70 dB SPL
(approximately 65.4 dBA). Speech stimuli in the current study were presented to
subjects at approximately 60 dBA. In Figure 7.6, 5.4 dB was subtracted from the
LTASS of Byrne et al. (1994) to provide an estimate of audibility for speech at
60 dBA. The upper and lower boundaries of the shaded area were defined by
adding 12 dB and subtracting 18 dB from these average level values for each
one-third octave frequency band.
All subjects showed thresholds that were within or below the speech banana
across the frequency range.
Generally, the high frequencies of 3000 and 4000 Hz were the least audible for
subjects with only 2 subjects (S06, S08) obtaining aided thresholds below the
speech banana at these frequencies. Most subjects showed the greatest
audibility in the low frequencies, with some subjects’ (S02, S03, S06, S08)
aided thresholds falling well below the speech banana at 250 and 500 Hz.
Percentage correct scores for the consonant test across the four low-pass filter
conditions are also shown for each subject in Figure 7.6. Error bars represent
plus and minus one standard deviation from the mean. Most subjects’ scores
appear to improve with increasing bandwidth regardless of the possible
presence or absence of dead regions. The exception is S06, whose scores
showed no improvement from the 2000 Hz bandwidth upwards. S06 was one of
the few subjects who showed elevated masked thresholds with the TEN test at
more than one frequency (See Figure 7.5). The consonant scores and TEN test
results appear to be consistent for this subject.
An analysis of variance (ANOVA) was carried out on subjects’ raw scores using
three factors. The factors were the speaker (male and female), vowel (/a/ and
/i/), and filter cut-off frequency (1400, 2000, 2800, and 5600 Hz).
- 115 0
100
20
80
40
60
60
40
80
100
S02
S01
120
0
20
0
100
20
80
40
60
60
40
80
S03
S04
120
0
20
0
100
20
80
40
60
60
40
80
100
20
S05
S06
120
0
0
100
20
80
40
60
60
40
80
20
S07
100
S08
120
0
20
0
100
80
40
60
60
40
80
100
120
S09
0.25
0.5
1
2
3 4
6
S10
0.25
0.5
1
2
3 4
6
20
0
Frequency (kHz)
Figure 7.6 Consonant test mean scores (filled circles) for S01-S10 are indicated on the
right y-axis as percentage correct. Error bars on each symbol represent plus and minus
one standard deviation from the mean. The speech banana for a speech input measured
at 60 dBA is shown by the shaded area. Aided thresholds (solid line) in dB SPL are
shown on the left y-axis.
Consonants correctly recognized (%)
Threshold (dB SPL)
100
- 116 – Chapter 7 Benefits of audibility for listeners with severe high-frequency hearing loss
Table 7.4 shows the F and p values for each factor and for the various
interactions. A three-way table of means is shown in Table 7.5. For factors
which varied within a session (4 filter conditions) a difference of greater than the
least significant difference (LSD) of 4.26 percentage points within a row of the
table represents a statistically significant difference (at the 5% level). For factors
which varied between sessions (2 speakers × 2 vowels) a difference of greater
than the LSD of 7.16 percentage points within a column of the table represents
a statistically significant difference.
It is clear from this analysis that consonant scores generally improved with
increasing bandwidth for the subject group. This score increase occurred with
both speakers and with both vowel conditions. The most notable score
increases were from 1400 Hz to 2000 Hz, and 2000 Hz to 5600 Hz. There was
no significant score increase from 2000 Hz to 2800 Hz. Subjects’ scores were
significantly lower with the vowel /i/ for all filter conditions and for both speakers.
Scores were not significantly different between the two speakers.
Factor/s
F value
p value
speaker
2.57
0.12
vowel
40.87
<0.001
filter
248.49
<0.001
speaker*vowel
1.12
0.3
speaker*filter
4.94
0.002
vowel*filter
1.88
0.134
vowel*filter*speaker
5.22
0.002
Table 7.4 Results of the Analysis of Variance for the consonant recognition test. Asterisk
symbols indicate interactions among factors.
In summary, average consonant scores for the subject group improved with
increasing audibility of high-frequency components of the speech signal. One
subject (S06) showed no improvement in scores with increasing audibility.
Scores were unaffected by the sex of the speaker. Subjects obtained higher
scores for the vowel /a/ than for the vowel /i/.
- 117 Speaker
Vowel
Female
/a/
/i/
/a/
/i/
Male
Low-pass filter cut-off frequency (Hz)
1400
2000
2800
5600
42.81
65.78
66.09
72.97
34.22
49.38
46.56
53.59
45.31
63.75
67.03
76.25
31.25
54.06
58.28
62.97
Table 7.5 Three-way table of means of subjects’ results. The mean percentage correct
score is shown for each combination of speaker, vowel and low-pass filter condition
used in the consonant test. A difference of greater than 4.26 percentage points within a
row represents a statistically significant difference at the 5 % level between two
conditions. A difference of greater than 7.16 percentage points across rows represents a
statistically significant difference at the 5 % level between two conditions.
7.5
Discussion
Previous publications have reported that for a small proportion of hearingimpaired listeners additional benefits in recognition are not seen when the
audibility of high frequencies is increased (Hogan et al., 1998; Murray et al.,
1986; Rankovic, 1991; Vickers et al., 2001). However, many of those studies,
found it difficult to provide a high degree of audibility of the speech signal. It has
been difficult, therefore, to separate the effects of limited signal audibility from
the effects of cochlear dead regions that may have been present in some of the
subjects who participated in the studies. The current study attempted to test
under conditions in which the speech signals were highly audible to subjects.
The average high-frequency gain at 1000, 2000, and 4000 Hz was calculated
for Experiment 1 to equal 54 dB (2cc coupler) for the subject group. In
comparison, the NAL-NL1 (Byrne et al., 2001) fitting guideline prescribes an
average of 40 dB gain (2cc coupler) at the same high frequencies for this group
of subjects.
On average, these subjects showed a significant score increase with increasing
bandwidth. Nine out of the ten subjects tested demonstrated a score increase
with increasing bandwidth. One subject showed no increases in speech scores
with increasing bandwidth beyond 2000 Hz. None of the subjects tested showed
a score decrease with increasing high-frequency audibility. The results seem to
indicate that whilst a worsening of performance is certainly possible when
amplifying high frequencies, it seems to occur in a very small number of cases,
and wasn’t observed in the current study.
The question then remains of how to identify these possible cases. One method
is speech recognition testing. This method is well known clinically and has the
- 118 – Chapter 7 Benefits of audibility for listeners with severe high-frequency hearing loss
advantage of using speech stimuli directly. An additional method that has been
suggested is the use of the TEN test, developed by Moore et al. (2000a). Only 3
subjects (S01, S02, S06) showed dead regions according to the TEN test. This
result could partly be due to limitations of the TEN test used in the current
study. Even the highest noise level was not sufficiently intense to provide
sufficient masking in the high frequencies. A new version of the test was
released in 2003 (Moore et al., 2003) which has some advantages over the
version used for Experiment 1. The bandwidth of the masking noise has been
reduced which allows for the noise to be increased in level, without the listener
experiencing loudness discomfort.
In cases where dead regions cannot be measured with the TEN test, they are
assumed to exist if the listener has pure-tone thresholds of 90 dB HL or greater
(Moore, 2001). Figure 7.1 shows the air-conduction thresholds measured for all
subjects. Five subjects (S03, S04, S05, S07, S09) had high-frequency
thresholds that were greater than or equal to 90 dB HL. A threshold greater than
or equal to 90 dB HL was measured for S03 at 3000, 4000, and 8000 Hz; for
S05 and S07, at 4000 and 8000 Hz; for S09 at 4000 Hz; and for S04 at 8000 Hz
only. These subjects had possible high-frequency dead regions based on the
severity of their hearing thresholds.
One subject (S06) showed speech recognition results in the consonant test that
were consistent with the presence of dead regions found in the TEN test.
Performance for this subject improved with increasing cut-off frequency until the
cut-off frequency was above the estimated edge frequency of the dead region.
Further increases in the cut-off frequency resulted in no further increases in
speech recognition score. The remaining nine subjects showed an increase in
scores with increasing bandwidth.
Both Vickers et al. (2001) and Moore (2001) conclude that there is little benefit
in amplifying frequencies well inside a dead region if that dead region extends
from its estimated edge frequency to the upper limit of hearing. However, there
may be some benefit in amplifying frequencies up to approximately one octave
above the estimated edge frequency of a dead region. The highest bandwidth
used in the current study (5600 Hz) was less than one octave above the
estimated edge frequency of the assumed dead region for subjects S01 and
- 119 S02 who both had dead regions according to the TEN test, as well as for those
subjects (S03, S04, S05, S07, S09) with assumed dead regions at high
frequencies based on their hearing thresholds being greater than 90 dB HL. In
addition, the dead regions found for S01 and S02 did not extend to the upper
limit of these subjects’ hearing.
It is possible that subjects’ speech scores may not have improved further if
additional higher cut-off frequencies had been tested. This would be difficult to
evaluate, as it would not be possible to achieve full audibility of the speech
stimuli beyond 5600 Hz given that stimuli were presented via a hearing aid. A
speech signal with very high gains would result in saturation of the hearing aid’s
output. It would be difficult to determine whether a flattening of speech
recognition scores with further increases of bandwidth was due to the presence
of a dead region, limited audibility of the speech signal, or some other cause
(such as distortion of the hearing aid). To determine which phonetic features
were affected most by bandwidth, the responses of each subject were analyzed
to determine the percentage information transmission for the phonetic features
of voicing, place, and manner (Miller et al., 1955) as a function of filter cut-off
frequency. The average results are shown in Figure 7.7.
Percent information transmitted
100
90
80
70
60
50
40
30
Voicing
Manner
Place
20
10
0
1400
2000
2800
5600
Frequency (Hz)
Figure 7.7 Average information transmission scores. The features of manner, place, and
voicing are plotted as a function of the low-pass filter cutoff frequency. Error bars
represent one standard deviation from the mean.
- 120 – Chapter 7 Benefits of audibility for listeners with severe high-frequency hearing loss
7.5.1 Conclusion
Extensive dead regions were found with the TEN test for one of
the ten subjects. For that subject, speech recognition scores did
not improve with increasing bandwidth. It is likely that many of
the remaining subjects also had dead regions above 3000 Hz
based on their hearing thresholds, but these could not be
demonstrated by means of the TEN test. On average, these
subjects showed a significant score increase with increasing
bandwidth. None of the subjects tested showed a score
decrease with increasing high-frequency audibility. Before
considering reducing high frequency amplification when fitting a
hearing aid, the clinician should consider whether the listener
demonstrated extensive dead regions that extended to the
upper limit of hearing. In addition, when a dead region is found
with the TEN test, it is important that amplification is provided to
at least one octave above the edge frequency of the dead
region (Baer et al., 2002; Vickers et al., 2001).
- 121 -
Chapter 8
Experiment 2
Frequency compression outcomes for listeners with steeplysloping losses
8.1
Introduction
Many listeners with severe high-frequency losses are able to make some use of
high-frequency speech cues if these cues can be made audible. However, the
levels of audibility achieved in Experiment 1 at these frequencies may be
impractical for everyday use of a hearing aid, and could result in discomfort for
the listener, and/or acoustic feedback. Alternative signal processing, such as
frequency compression or cochlear implantation, may be able to provide these
listeners with additional high-frequency information while avoiding these
problems.
This chapter describes the methods followed (section 8.2) and results found
(section 8.3 and 8.4) when the frequency-compression scheme implemented by
Simpson et al. (2005c) was fitted to a group of 7 listeners with steeply-sloping
losses. Subjects were initially fitted with conventional hearing aids, and a
comparison was made between these conventional aids and the subjects’ own
hearing aids. Those subjects who obtained improved or equivalent speech
perception with the conventional aids were fitted with the frequencycompression scheme. They were required to wear the frequency-compression
device at home for 4 -5 weeks. Speech recognition testing was carried out to
determine what effect the frequency-compression scheme had on each
subject’s perceptual abilities. The results of this experiment will be published
shortly (Simpson et al. 2006). A copy of the publication can be seen in Appendix
B.
8.2
Methods and Procedures
An outline of the testing carried out for each subject is shown in Figure 8.2.
Subjects were fitted with a conventional device and the experimental frequencycompression device. Speech perception testing was carried out with the
- 122 – Chapter 8 Frequency compression outcomes
subjects’ own hearing aids, the conventional device, and the frequencycompression device. The methods followed when fitting devices and carrying
out speech testing are described in detail below. It was anticipated at the time of
conducting Experiment 2 that some subjects would decide to undergo cochlear
implantation at a later date. It was unknown at the time of conducting
Experiment 2 which ear would be chosen for implantation as well as what effect
implantation would have on hearing thresholds and speech perception. For this
reason, testing was carried out in the binaural condition as well as with the two
ears separately.
8.2.1 Subjects
Seven hearing-impaired adults, comprising 3 women and 4 men, participated in
the study. Four of the subjects were experienced hearing-aid users, whereas
three of the subjects had not worn hearing aids previously.
8.2.1.1 Criteria for recruitment
The criteria considered when approaching potential participants were developed
with the aim of recruiting subjects with the following characteristics:
1. Eighteen years of age or older.
2. Severe-to-profound sensorineural hearing loss for frequencies > 2000 Hz, but
with bilateral normal hearing thresholds, or a mild-to-moderate hearing loss, for
frequencies ≤1000 Hz. Thresholds for mid-frequencies would fall within the
mild-through-profound range.
3. Postlinguistic onset of bilateral severe-to-profound sensorineural hearing
loss.
4. English spoken as a primary language.
Relevant information about the subjects is provided in Table 8.1. Their hearing
threshold levels, measured conventionally under headphones, are shown in
Figure 8.1. For all subjects, hearing losses were assumed to have primarily a
sensorineural origin, based on standard air- and bone-conduction audiometry.
Subjects were not tested for cochlear dead regions. It was assumed that the
TEN would not be intense enough to mask effectively at all frequencies.
- 123 However, as discussed later, the audiogram configurations suggest that dead
regions were present at high frequencies in most, if not all ears.
Subject
Age
S32
75
S34
Sex
Probable
etiology of
hearing loss
Type of
Own Aid/s
M
Unknown
Widex
Senso Diva
Digital, adaptive
directional
microphone
Compression
33
F
Unknown
No hearing
aid used
previously
-
-
S35
33
M
Noise
Exposure
No hearing
aid used
previously
-
-
S36
57
M
Hereditary
No hearing
aid used
previously
-
-
S37
53
M
Hereditary
Miracle Ear
Linear
S38
68
F
Unknown
Siemens
Analog,
omnidirectional
Digital, adaptive
directional
microphone
S39
72
F
Ototoxic
drugs
Bernafon
RB15
Digitally
programmable,
omnidirectional
Linear
S41
33
F
Unknown
Phonak
PPCL4
Analog
Linear
S42
67
F
Unknown
Siemens
Prisma
Digital, adaptive
directional
microphone
Compression
Features of
Own Aid(s)
Processing
Strategy of
Own Aid(s)
Compression
Table 8.1 Relevant information about the subjects (S32 – S39) who participated in
Experiment 2, and their hearing aids. (S41 and S42 participated in Experiment 3).
8.2.1.2 Ethical approval
This study was approved by the Human Research and Ethics Committee at the
Royal Victorian Eye and Ear Hospital (Project No 99 / 362H). The test
procedure and aims of the study were discussed with each subject. Written
information regarding the study was given to each subject (Appendix A).
Informed consent was obtained from each subject. Subjects were not paid for
their participation in the experiment, although expenses such as travel costs
were reimbursed.
- 124 – Chapter 8 Frequency compression outcomes
0
20
40
60
80
S32
S35
Left ear
Right ear
100
120
0
20
40
Hearing Level (dB HL)
60
80
S34
S36
100
120
0
20
40
60
80
S37
S39
100
120
0
0.125
0.25
0.5
1 1.25
4
8
20
40
60
80
100
S38
120
0.125
0.25
0.5
1
1.6
4
8
Frequency (kHz)
Figure 8.1 Hearing threshold levels (dB HL) for the subjects who participated in
Experiment 2. Left ear thresholds for each subject are shown by the solid line.
Thresholds obtained for the right ear are shown by the dotted line. The hatched area
represents frequencies which were compressed. The lower edge of the hatched area
represents the final cut-off frequency for each subject. Thresholds of subjects who were
fitted with a cut-off frequency of 1.6 kHz are shown on the left of the figure, and those of
subjects who were fitted with a cut-off frequency of 1.25 kHz are shown on the right.
Figure from Simpson et al. (2006).
- 125 -
OWN AIDS/
NO AIDS
TEST CONDITIONS
LEFT EAR
ALONE
RIGHT EAR
ALONE
BINAURAL
CNC words
CNC words
Consonants
TIME
CONVENTIONAL
DEVICE
Conventional device fitting
Sentences in noise
FREQUENCYCOMPRESSION
DEVICE
Frequency-compression fitting
CNC words
Consonants
Sentences in noise
APHAB questionnaire
Figure 8.2 Flow chart of the testing carried out for Experiment 2 for each of the 7 subjects
who participated in the study.
- 126 – Chapter 8 Frequency compression outcomes
8.2.2 Own-aid testing
8.2.2.1 Word recognition in quiet
Stimuli
Consonant – vowel nucleus – consonant (CNC) monosyllabic word lists
(Peterson et al., 1962) was presented from audio recordings. There were 50
words per list, spoken by a male with a typical Australian accent. No lists of
words (other than practice lists) were repeated for any subject during the trial.
The order in which lists were presented to subjects across sessions was
randomized. The average level of the words, when measured at the subject’s
listening position, was 60 – 65 dBA. These levels, which are similar to the levels
corresponding to between conversational and raised speech, were generally
perceived as comfortably loud when heard by the subjects through their hearing
aids.
Method
The hearing-aid usage and medical history of each subject was documented
during the first test session. A pure-tone audiogram, including both air and bone
conduction, was obtained, and the electro-acoustic characteristics of each
subject’s own hearing aids were measured and recorded. Table 8.1 includes
relevant details of each subject’s own aids.
For all evaluations of speech intelligibility, each subject was tested individually
in a medium sized sound-attenuating booth. Three of the subjects (S34, S35,
S36) did not wear any hearing aids at the time of testing. For these subjects,
testing was carried out unaided. For the remaining four subjects, the volume
control on each subject’s own hearing device was set for comfortable listening
to speech at a conversational level in quiet conditions. For most subjects, this
was the default volume control setting. This setting was noted and fixed for all
following test sessions.
The hearing aid in the non-test ear was removed and switched off during
testing. In addition, the non-test ear was masked. A Madsen Aurical audiometer
was used to generate a speech-weighted masking noise, which was presented
to the subject under Telephonics TDH39 headphones. The amount of masking
was determined individually for each subject according to the following formula:
- 127 M = P – 40 + G +10 + C
Equation 5
Where M was the speech masking dial reading, P was the presentation level of
speech (in dB SPL), G was the air-bone gap in the non-test ear, and C was the
audiometer conversion from dB SPL to dial reading. Forty was subtracted for
the interaural attenuation effects for speech. Ten was added to ensure that the
spectral peaks of the speech signal were masked during testing.
A practice CNC word list was presented to familiarize subjects with the testing
procedure and materials. Subjects were instructed to repeat each word
immediately after hearing it, and to guess if unsure. Responses from the
practice list were excluded from the data analysis. After the practice list, a
further two lists were carried out in each of the following two conditions: (1) leftear alone, and (2) right-ear alone. Subjects’ responses were analyzed to
determine the number of phonemes correctly recognized out of a total of 150
phonemes per list.
8.2.3 Conventional device fitting and testing
8.2.3.1 Fitting method
Each subject was fitted with identical conventional hearing aids. These hearing
aids (Phonak Supero 412) were the same as those used for Experiment 1. The
conventional hearing instruments were fitted to each subject using appropriate
fitting software (Phonak Fitting Guideline version 8.1), with which userselectable normal and noise-reduction programs were created. The subjects’
pure-tone thresholds were entered into the fitting software to derive an initial
fitting suggestion based on the manufacturer’s recommendation. This fitting
suggestion was derived from the National Acoustic Laboratories non-linear
fitting rule (NAL-NL1) (Byrne et al., 2001). When necessary, these settings were
altered slightly at the follow-up sessions based on subject feedback. For
example, the three most common subject reports regarded their own voice,
loudness discomfort, and general sound quality. Adjustments were made to the
low-frequency gain of the device in steps of 3 dB if the subject was dissatisfied
with the sound of their own voice. If the subject reported discomfort when
- 128 – Chapter 8 Frequency compression outcomes
listening to loud noises, the overall maximum power output of the hearing aid
was reduced in steps of 3 dB. Typically, changes of no greater than two steps
were applied when these adjustments were required. If the subject reported
dissatisfaction with the general sound quality of the device and they were a
previous hearing-aid user, the programming of the device was adjusted to
approximate the amplification characteristics of that subject’s own hearing aids,
based on 2-cm3 coupler measurements.
8.2.3.2 Speech test stimuli and methods
Speech perception comparisons were made between the conventional hearing
device and the subjects’ own hearing aids for those subjects who had worn
hearing aids previously. Subjects were selected to be fitted with the frequencycompression device only if their speech perception results with the conventional
hearing device were approximately the same as, or better than, those obtained
with their own hearing aids. Each subject had been wearing the conventional
hearing devices for 4 – 5 weeks prior to the commencement of wearing the
frequency-compression device. Speech perception testing was carried out
during the final two testing sessions of wearing the conventional device.
The speech tests comprised recognition of open-set monosyllabic words,
closed-set medial consonants in quiet, and open-set sentences presented with
a competing noise. For all speech testing where the two ears were tested
separately the non-test hearing aid was removed and switched off during
testing. The non-test ear was masked with a speech-weighted noise during
testing following the same specifications as those described in section 8.2.2.1
(Method).
8.2.3.2.1 Word recognition in quiet
Two lists of CNC words (described in section 8.2.2.1 Method) were presented at
60 - 65 dBA in each of the following conditions: right hearing-aid alone, left
hearing-aid alone, and bilaterally.
- 129 8.2.3.2.2 Consonant recognition in quiet
Stimuli
Vowel-consonant-vowel (VCV) utterances recorded by a male speaker were
used for this test. These were the same stimuli as those used in Experiment 1
(described in Chapter 7, section 7.3.5 Stimuli). The 16 consonants selected for
the test were: /p/, /t/, /k/, /b/, /d/, /g/, /m/, /n/, /s/, /S/, /z/, /f/, /v/, /tS/, /j/, and /T/.
The initial vowel was always the same as the final vowel.
The levels of the medial consonant tokens were approximately equalized as
described in Chapter 7 (see section 7.3.5 Stimuli). The average levels were
approximately 60 dBA at 1 m from the loudspeaker. Each stimulus presentation
was repeated six times making a total of 96 tokens. Stimuli were presented in a
random order.
Method
The consonant test was also carried out in the following conditions: binaurally,
left hearing-aid alone, and right hearing-aid alone. As described in Chapter 7
(see section 7.3.5 Method) a closed-set procedure was used for the test of
consonant recognition, in which 16 buttons appeared on a computer screen.
Figure 7.4 shows how the test screen appeared to subjects. The buttons were
of the following format: vCv, where the vowel was /a/, and C represented each
consonant. The test consisted of three blocks in which each token was heard
twice by the subject. The stimuli were presented in a random order in each
subtest. The subjects were presented with the 16 tokens in the test set, as
displayed on the computer screen. They were instructed to identify which
consonant they heard after each stimulus was presented by pressing the
corresponding button on the screen. There was no option for a response other
than one of the 16 consonants in the test set. At the start of each test session,
each subject carried out one block to gain practice and confidence in the
procedure. The results were recorded as usual, but disregarded in the final
analysis of the data.
- 130 – Chapter 8 Frequency compression outcomes
8.2.3.2.3 Sentence recognition in noise
Stimuli
The material used was CUNY sentence lists (Boothroyd et al., 1985). There
were a total of 60 available lists. Each list consisted of 12 sentences (i.e.,
approximately 102 words). This material, which was recorded by a female
speaker with a typical Australian accent, was presented through the
loudspeaker at an average level of approximately 65 dBA. As described below,
an adaptive procedure was used to estimate the signal-to-noise ratio (SNR) for
a target score of 50% correct. For this procedure, the competing noise was
eight-talker babble. This noise was mixed at a controlled level with the sentence
material using a two-channel audiometer. As for CNC words (see section
8.2.2.1 Method), the CUNY sentence lists were selected at random, and no list
was repeated for any subject throughout the test sessions.
Method
Initially, one practice list was presented, and the subject was instructed to
repeat as many words as they could identify after hearing each sentence, and
to guess if unsure. A further four lists were then presented, and one point was
given for each word repeated correctly by the subject. The total was converted
to a percentage score. This score had to be equal to or greater than 75% for
testing to continue. It was assumed that subjects with scores of less than 75%
in quiet would not score above 50% with a competing noise source. Therefore,
subjects with scores of less than 75% in quiet were not tested further with
sentences in noise.
Subsequently, sentences were presented combined with noise starting at an
SNR of 10 dB. Although pauses were inserted between sentences to enable
scores to be calculated, the noise was presented continuously throughout the
testing procedure. After the subject’s response to each pair of consecutive
sentences (containing approximately 9-13 words), the proportion of words
correctly identified was determined. The level of the noise was either decreased
or increased, according to whether the subject’s score was respectively less or
greater than 50% correct. Initially, changes in the noise level were made in
steps of 5 dB, and each change in the direction of noise-level adjustments was
- 131 noted. After two such reversals were found, subsequent changes in the noise
level were made in steps of 3 dB. The test continued until a further 10 reversals
were obtained. The final SNR for one test run was the average across the last
10 reversals.
8.2.4 Frequency-compression fitting and testing
The processing of sounds by the frequency-compression device was described
in detail in Chapter 4, section 4.4 (Speech perception results with a novel
frequency-shifting device). The hardware of the frequency-compression hearing
instrument was identical to that implemented in previous studies (Simpson et
al., 2005c). It consisted of two main parts: a pair of modified behind-the-ear
(BTE) conventional hearing devices, and a SHARP programmable body-worn
speech processor (Zakis et al., 2004). A schematic representation of the
hardware of the frequency-compression hearing instrument is shown in Figure
8.3. An in-house software program 4 was used to fit and make adjustments to
the device.
Left Microphone
Left Receiver
Left BTE
Low-pass filter
∑
Frequency
compression
processing
Right BTE
Right Microphone
Low-pass filter
∑
Sharp body-worn processor
Right Receiver
Figure 8.3 Block diagram of the binaural signal processing implemented in the frequency
compression hearing device. Figure reproduced from Simpson et al. (2005c).
For all speech testing where the two ears were tested separately, the non-test
hearing aid was removed and muted during testing. The non-test ear was
4
The software program was developed by Hugh McDermott, Adam Hersbach and Rodney
Millard at the University of Melbourne.
- 132 – Chapter 8 Frequency compression outcomes
masked with a speech-weighted noise during testing following the same
specifications as those described in section 8.2.2.1 (Word recognition in quiet,
Method).
8.2.4.1 Device fitting stimuli and method
Fitting stimuli
The rationale for fitting the frequency-compression scheme was to ensure that
speech (when heard through the device) would be both audible and comfortable
for subjects. For this reason, equal loudness level measurements were used to
fit the SHARP frequency-compression device. Third-octave narrowband noises
with duration of 0.5 sec were selected. Each noise stimulus was separated from
the next stimulus by a silent interval of duration 0.5 sec. The noises were
ramped in level at each end with linear ramps of duration 30 ms. The level of
each noise was set to equal the average one-third octave level at the
corresponding frequency of the LTASS (Byrne et al., 1994) for an overall level
of 70 dB SPL measured at the subject’s listening position.
Fitting method
Initially, an informal listening task was carried out to determine whether subjects
with steeply-sloping losses would prefer certain frequency-compression
parameters. There are two main parameters which can be adjusted in the
device. The first is the point at which frequency-compression begins, or the cutoff frequency. The second parameter is the degree of frequency compression,
or the slope of the input-frequency to output-frequency relationship. Both of
these parameters were adjusted in a variety of ways in an attempt to determine
an acceptable fitting for this group of subjects. The parameters selected to be
investigated further were cut-off frequencies of 1000, 1250, and 1600 Hz and
compression slopes of 4:1, 2:1, and 0.5:1.
Ten hearing-impaired adults with steeply-sloping losses participated in a pilot
study. A paired comparison method was used to determine which settings
subjects preferred. Of the 10 subjects, only four consistently chose the same
parameter combination. In general, it appeared that subjects preferred the
higher cut-off frequencies of 1250 and 1600 Hz, combined with an input-tooutput frequency compression slope of 2:1. These preferred parameters (cut-off
- 133 frequency of 1250 or 1600 Hz, and a slope of 2:1) were then fixed when fitting
the frequency-compression hearing aid, as detailed below.
The frequency-compression device was fitted in a similar manner to that
described in Simpson et al. (2005c). Each subject’s fitting parameters
determined as described in section 8.2.3.1 (Fitting Method) were programmed
into modified conventional hearing devices attached to the SHARP processor.
The loudness of the test signals specified above was approximately equalized
across frequency. To achieve this, amplification of the high-frequency test
signals (described in section 8.2.4.1) was adjusted by means of a loudness
rating method.
The subject was seated, wearing the frequency-compression device, at a
distance of 1 m directly in front of the loudspeaker. Subjects were given a
categorical list of 9 loudness levels (Hawkins et al., 1987). The categories were:
very soft, soft, comfortable but slightly soft, comfortable, comfortable but slightly
loud, loud but OK, uncomfortably loud, extremely uncomfortable, and painfully
loud. The subject was asked to indicate the loudness of the noise stimuli by
pointing to a category on the printed list. Frequency compression was enabled
in the device before presenting narrow-band noises at 1000, 1250, 1600, 2000,
2500, 3100, and 4000 Hz. Subjects were thus making a loudness judgment
based on the compressed signal. The amplification of each frequency was
adjusted manually by the clinician for each narrow-band noise until the subject
reported each noise to be “comfortable but slightly soft”. Due to the severity of
hearing loss for most subjects at these frequencies, it was often not possible to
achieve this level at all frequencies without feedback oscillation occurring. In
these cases, the level was set to the maximum possible without feedback
oscillation. Initially a cut-off frequency of 1250 Hz was chosen for each subject.
After fitting, subjects were asked to wear the device away from the laboratory
and provide feedback about it at two follow-up sessions. If the subject was
dissatisfied with the sound quality of the device, the output level and the cut-off
frequency were adjusted at these sessions. The clinician first reduced the level
of the frequency-compressed signal by 3-5 dB. If the subject continued to report
an unacceptable sound quality, then the cut-off frequency was adjusted to 1600
Hz. Figure 8.1 shows the final cut-off frequencies selected for each subject.
- 134 – Chapter 8 Frequency compression outcomes
Three subjects (S32, S34, S39) were fitted with a cut-off frequency of 1250 Hz,
and four subjects (S35, S36, S37, S38) were fitted with a cut-off frequency of
1600 Hz. As mentioned above, the same frequency-compression slope was
selected for all subjects. No further adjustments were made to the program for
the remainder of the trial.
8.2.4.2 Speech test stimuli and method
Identical speech test stimuli to those described for the conventional aid fitting
and testing were used (see section 8.2.3.2 Speech test stimuli and methods).
An identical procedure was used for carrying out speech testing to that
described under the conventional device fitting and testing (see section 8.2.3.2
Speech test stimuli and methods).
8.2.4.3 Subjective assessment
The Abbreviated Profile of Hearing Aid Benefit (Cox et al., 1995) was used to
measure subjective benefit for the conventional and experimental devices. The
APHAB is a 24-item self-assessment scored on four subscales (6 items per
subscale). Three subscales, Ease of Communication, Reverberation, and
Background Noise address speech understanding in everyday life. The fourth
subscale, Aversiveness to Sounds, measures negative reactions to
environmental sounds. The APHAB was given to subjects on the second-last
testing session while wearing the frequency-compression device.
8.3
Results: own aid versus conventional device
A comparison was made between subjects’ own hearing aids and the
conventional device for those subjects who had worn hearing aids previously to
determine which subjects would take part in the frequency-compression trial.
Subjects were divided into two groups: those who had worn hearing aids
previously (S32, S37, S38, S39, S41, S42), and those who had not worn
hearing aids before (S34, S35, S36). S41 and S42 did not participate in the
frequency-compression trial. Both these subjects were recruited at a later date
in the study. Initial results at the time of their recruitment indicated that the
experimental frequency-compression scheme would most likely not have
provided these two subjects with perceptual benefits. However, both subjects
were interested in cochlear implantation. As such, S41 and S42 were included
- 135 in Experiment 3, described in Chapter 9. Their results when their own aids were
compared with conventional aids are shown below.
8.3.1 Word recognition in quiet
Figure 8.4 shows the mean percentage correct phoneme scores obtained by
the six subjects (S32, S37, S38, S39, S41, S42) who were experienced
hearing-aid users who participated in the study. Subjects’ scores when wearing
their own hearing aids (OA) were compared with the scores obtained with the
conventional hearing device (CD). S42 did not like the sound quality of the
conventional device and did not wear it. Therefore, her scores are only shown
for the OA condition. S42’s data was not included in statistical analysis. A twofactor analysis of variance (ANOVA) was carried out on the remaining five
subjects’ scores. Across all subjects there was no statistically significant
difference (p = 0.79, df = 1) between the subjects’ own aids and the
conventional hearing device for the right ear. No significant interaction was
found between the scheme and the subject factors (p = 0.4).
However, speech recognition scores across the subject group were improved
significantly (p < 0.001, df = 1) with the conventional device when compared
with the subjects’ own aids for the left ear. A significant interaction was found
between the scheme and the subject factors (p = 0.01). Therefore, each
subject’s data was analyzed further separately with pair-wise comparisons
using the Holm-Sidak method (Hochberg et al., 1987). Although visual
inspection shows higher individual scores with the conventional device when
compared with own aids for all subjects, this difference was significant only for
S37 (t = 2.511, p = 0.03) and S41 (t = 6.847, p < 0.001).
The mean percentage correct scores for the three subjects (S34, S35, S36)
who were non-hearing aid users can be seen in Figure 8.5. Subjects’ scores
obtained unaided were compared with the scores obtained with the
conventional hearing device (CD). A two-factor ANOVA was carried out on all
subjects’ mean scores. No statistically significant difference (p = 0.39, df = 1)
between subjects’ unaided scores and the conventional hearing device for the
left ear was found across the subject group. However, speech recognition
scores were improved significantly (p = 0.01, df = 1) with the conventional
device when compared to subjects’ unaided scores for the right ear. No
- 136 – Chapter 8 Frequency compression outcomes
significant interaction was found between the scheme and the subject factors
(p = 0.1) for both the left and right ear conditions. All subjects performed the
same as, or better, with the conventional device when compared to their own
aids when phoneme scores were analyzed at a moderate level in quiet.
Phonemes correctly recognized (%)
Therefore, all subjects were included in the frequency-compression trial.
100
Own Aid (Right ear)
Conventional device (Right ear)
Own aid (Left ear)
Conventional device (Left ear)
90
80
70
***
60
*
50
40
***
30
20
10
0
S32
S37
S38
S39
S41
S42 Mean
Figure 8.4 Mean phoneme scores for the left and right ears when listening to
monosyllabic words obtained by the experienced hearing-aid users who participated in
the study. Unfilled columns show scores subjects obtained with the right ear when using
their own hearing aids (OA), filled red columns show scores obtained with the right ear
when using the conventional hearing devices (CD), filled grey columns show scores
subjects obtained with the left ear when using their own hearing aids, and filled blue
columns show scores obtained with the left ear when using the conventional hearing
devices (CD). Scores averaged across subjects are shown in the rightmost columns, with
error bars indicating one standard deviation. Statistical significance is shown by asterisk
symbols: * 0.01 < p < 0.05, ** 0.001 < p < 0.01, *** p < 0.001.
Phonemes correctly recognized (%)
- 137 -
100
Own aid (Right ear)
Conventional device (Right ear)
Own aid (Left ear)
Conventional device (Left ear)
90
80
70
**
60
50
40
30
20
10
0
S34
S35
S36
Mean
Figure 8.5 Mean phoneme scores for the left and right ears when listening to
monosyllabic words obtained by the non-hearing-aid users who participated in the study.
Unfilled columns show scores subjects obtained with the right ear when using their own
hearing aids (OA), filled red columns show scores obtained with the right ear when using
the conventional hearing devices (CD), filled grey columns show scores subjects
obtained with the left ear when using their own hearing aids, and filled blue columns
show scores obtained with the left ear when using the conventional hearing devices
(CD). Scores averaged across subjects are shown in the pair of rightmost columns, with
error bars indicating one standard deviation. Statistical significance is shown by asterisk
symbols: * 0.01 < p < 0.05, ** 0.001 < p < 0.01, *** p < 0.001.
Results: conventional device versus frequency
compression
8.4
8.4.1 Word recognition in quiet
For the CNC word tests, the mean phoneme scores obtained by each subject
with their conventional hearing devices and with the experimental device for the
binaural condition, left-ear alone, and right-ear alone are shown in Figure 8.6.
Across the 7 subjects, speech recognition scores were compared with the two
hearing-aid processing schemes by means of a two-factor ANOVA. No
statistically significant difference was found for the frequency-compression
scheme (FrC) when compared to the conventional hearing device (CD) in any of
- 138 – Chapter 8 Frequency compression outcomes
the three conditions tested [binaural (df = 1, p = 0.11), left-ear alone (df = 1, p =
0.07), right-ear alone (df = 1, p = 0.07)]. In addition, there was no significant
interaction between the scheme and subject factors in any of the three
conditions [binaural (p = 0.8), right-ear alone (p = 0.25), left-ear alone (p =
Phonemes correctly recognized (%)
0.59)].
100
Binaural CD
Binaural FrC
90
Left ear CD
Left ear FrC
Right ear CD
Right ear FrC
80
70
60
50
40
30
20
10
0
S32
S34
S35
S36
S37
S38
S39 Mean
Figure 8.6 Mean phoneme scores for the binaural, left-ear alone, and right-ear alone
conditions for the 7 subjects who participated in Experiment 2 when listening to
monosyllabic words. For the binaural condition, filled grey hatched columns show
scores subjects obtained with the frequency compression device (FrC), and unfilled
columns show scores obtained using the conventional hearing devices (CD). For the leftear alone, filled blue-hatched columns show scores subjects obtained with FrC, and filled
blue columns show scores obtained using the CD. For the right-ear alone, filled redhatched columns show scores subjects obtained with FrC, and filled red columns show
scores obtained using the CD. Scores averaged across subjects are shown in the pair of
rightmost columns, with error bars indicating one standard deviation.
8.4.2 Consonant recognition in quiet
The mean percentage correct scores obtained by each subject comparing the
conventional hearing device (CD) and the experimental device (FrC) in all 3
conditions (binaural, left-ear alone, and right-ear alone) are shown in Figure 8.7.
A two-way analysis of variance was carried out on all subjects’ mean scores.
Across all subjects there was no statistically significant difference between the
conventional hearing device and the experimental device for the binaural and
- 139 right-ear alone conditions [binaural (df = 1, p = 0.19), right ear alone (df = 1, p =
0.5)]. There was a significant difference in scores across the subject group for
the left-ear alone (df = 1, p < 0.001).
A significant interaction between the scheme and subject factors was found for
all three conditions [binaural (p = 0.002), right-ear alone (p = 0.01), left-ear
alone (p = 0.002)]. Each subject’s data was analyzed further separately with
pair-wise comparisons (Holm-Sidak method). In the binaural condition, speech
recognition scores for individual subjects were improved with the experimental
device when compared with the conventional hearing device for S35, S36, and
S37 but the score increase was significant for S37 (t = 2.736, p = 0.011) only.
Subjects S32, S34, and S38 showed a decrease in scores with the
experimental device when compared to the conventional hearing device. This
decrease was significant for subjects S32 (t = 3.719, p < 0.001) and S34 (t =
2.177, p = 0.038).
For the left-ear alone, subjects S32, S35, S36, S37, and S38 showed a score
increase with the experimental device. This score increase was significant for
S35 (t = 4.744, p < 0.001) and S37 (t = 4.111, p < 0.001). One subject (S34)
obtained a score decrease with the device. However, this was not found to be
significant.
Subjects S32, S34, S35, and S36 obtained an improvement in speech scores
with the experimental device for the right-ear alone, and two subjects (S37,
S39) showed a decrease in speech scores with the device. This score
difference was shown to be significant for S32 (t = 2.909, p = 0.007) and S39 (t
= 3.098, p = 0.004) only.
These results show a large variability among the subjects with a significant
difference across the subject group between the conventional device and the
experimental device found only for the left-ear alone condition. Of the 7
subjects, none showed a significant improvement with the experimental device
across all three conditions. One subject (S37) showed a significant
improvement in speech scores for the left-ear alone and binaural conditions.
None of the remaining subjects obtained a significant improvement in one ear,
together with an improvement in speech scores for the binaural condition. No
subject obtained significantly lower speech scores with the experimental device
- 140 – Chapter 8 Frequency compression outcomes
across all three conditions. None of the subjects obtained a significant score
decrease in one ear, together with a decrease in speech scores for the binaural
Consonants correctly recognized (%)
condition.
100
Binaural CD
90
Binaural FrC
Left CD
Left FrC
Right CD
Right FrC
80
70
60
*
***
50
40
30
***
**
***
***
**
**
20
10
0
S32
S34
S35
S36
S37
S38
S39 Mean
Figure 8.7 Mean consonant scores obtained by the 7 hearing-impaired subjects who
participated in Experiment 2. For the binaural condition, filled grey-hatched columns
show scores subjects obtained with the frequency-compression device (FrC), and
unfilled columns show scores obtained using the conventional hearing devices (CD). For
the left-ear alone, filled blue-hatched columns show scores subjects obtained with FrC,
and filled blue columns show scores obtained using the CD. For the right-ear alone, filled
red-hatched columns show scores subjects obtained with FrC, and filled red columns
show scores obtained using the CD. Scores averaged across subjects are shown in the
pair of rightmost columns, with error bars indicating one standard deviation. Statistical
significance is shown by asterisk symbols: * 0.01 < p < 0.05, ** 0.001 < p < 0.01, *** p <
0.001.
8.4.3 Sentence recognition in noise
The SNRs obtained by each subject from the adaptive test of sentence
comprehension in noise are shown in Figure 8.8. Note that larger SNR values
result from poorer performance on this test; therefore, shorter columns in the
graph indicate better performance. As indicated on the vertical axis, 5 subjects
participated in these tests. The excluded subjects were S32 and S39 who
- 141 scored less than 75% words correctly identified in the initial sentence test
without competing noise.
Mean
*
S38
S37
S36
***
S35
CD
FrC
S34
0
2
4
6
8
10
12
14
16
SNR (dB)
Figure 8.8 Mean signal-to-noise ratios (SNRs) obtained by the subjects. Unfilled columns
show SNRs obtained with the conventional device (CD), and filled columns show SNRs
obtained with frequency compression (FrC). SNRs averaged across subjects are shown
in the top pair of columns, with error bars indicating one standard deviation. Statistical
significance is shown by asterisk symbols: * 0.01 < p < 0.05, ** 0.001 < p < 0.01, *** p <
0.001.
Although some variability in the results from this test was found among the
subjects, a two-factor analysis of variance revealed that the score differences
were statistically significant (df = 1, t = 2.532, p = 0.02) across the group. A
significant interaction term (subject × scheme, p < 0.001) was also found. An
analysis of each subject’s data separately with pair-wise comparisons (HolmSidak method) found that one subject (S35) showed a significant improvement
(t = 5.233, p < 0.001) when wearing frequency compression. No significant
differences were found for the remaining four subjects. To investigate this
further, the ANOVA was repeated with S35’s scores omitted from the analysis.
In this analysis, no significant score differences were found (p = 0.85) across
the group, nor was there a significant interaction between subject and scheme
factors (p = 0.34).
- 142 – Chapter 8 Frequency compression outcomes
8.4.4 Subjective assessment
The APHAB questionnaire was scored according to the guidelines supplied by
Cox & Alexander (1995). Each of the 24 questions was assigned a score in one
of four subscales: ease of communication (EC), background noise (BGN),
reverberation (RV), and aversiveness (AV). A global score was also determined
by averaging the scores obtained from all four subscales. The total score
obtained in each subscale with frequency compression was subtracted from the
total score obtained with the conventional device. These values are shown for 6
subjects in Figure 8.9.
Prefers FrC
20
10
Prefers CD
0
-10
-20
EC
BN
RV
-30
AV
Global
-40
S34
S35
S36
S37
S38
S39
Figure 8.9 Preference scores from the APHAB questionnaire provided by 6 subjects who
participated in the study. For each subject, the subscale Ease of Communication (EC) is
shown by the filled black columns, the subscale Background Noise is shown by the filled
pink columns, the subscale Reverberation (RV) is shown by the unfilled columns, the
subscale Aversiveness (AV) is shown by the filled green columns, and Global scores are
shown by the filled grey columns. As shown on the vertical axis, negative values indicate
a preference for the conventional device (CD), whereas positive values indicate a
preference for the frequency-compression device (FrC).
One subject (S32) did not return the APHAB questionnaire. Positive values
indicate a preference for frequency compression, whereas negative values
indicate a preference for the conventional device. Preference ratings varied
across subjects and across subscales. The conventional device was preferred
- 143 by two subjects (S34, S35) for the subscale aversiveness, two subjects (S36,
S37) for the subscale ease of communication, and one subject (S37) for the
subscales of background noise and reverberation. The frequency-compression
device was preferred by two subjects (S34, S38) for the subscale background
noise, one subject (S34) for the subscale reverberation, and one subject (S36)
for the subscale of aversiveness. Globally, higher scores for the conventional
device are shown for four of the six subjects. Of the remaining two subjects,
S38 showed a higher score for the frequency-compression device whereas S39
showed no difference in scores between the two devices. A two-way ANOVA
was carried out on the subjects’ average scores. No significant difference was
found between the conventional device and the frequency-compression device
for any of the subscales [EC (df = 1, p = 0.18), BGN (df = 1, p = 0.56), RV (df =
1, p = 0.68), AV (df = 1, p = 0.36), Global (df = 1, p = 0.18)].
8.5
Discussion
The frequency-compression scheme provided perceptual performance similar,
on average, to the performance of conventionally-fitted hearing aids for words
and consonants presented at a moderate level in quiet conditions for the 7
subjects who participated in the study. For the CNC word test in all three
conditions (binaural, left-ear alone, and right-ear alone) scores for individual
subjects showed no significant differences between the conventional device and
the frequency-compression device.
For the test of consonants, there was no significant difference across the group
between the conventional device and the frequency-compression scheme in the
binaural and right-ear alone conditions. For the binaural condition, one subject
(S37) showed a significant score improvement, and two subjects (S32, S34)
showed a significant decrease when listening with the frequency-compression
device. For the right-ear alone condition, one subject (S32) showed a significant
score improvement, and one subject (S39) showed a significant score decrease
when listening with the frequency-compression device. A significant difference
across the group was found for the left-ear alone condition. Individual results
showed two subjects (S35, S37) with a significant score increase when listening
with the frequency-compression device. Results with the consonant test were
- 144 – Chapter 8 Frequency compression outcomes
variable across subjects and no subject showed a consistent score increase or
decrease with the frequency-compression scheme across all three conditions.
When tested with sentences in noise, the frequency-compression scheme
demonstrated a significant improvement for one of the 5 subjects tested. It is
possible that the frequency-compression device may have resulted in
perceptual improvements for certain sounds at the expense of others, resulting
in subjects’ overall percentage correct score remaining unaffected. To
investigate this further, confusion matrices were constructed from the subjects’
binaural consonant test responses for both the conventional hearing aids and
the experimental device. A confusion matrix provides information about how
phonemes have been identified by subjects. A “confusion” occurs if a particular
phoneme has been consistently incorrectly identified for another phoneme.
These matrices are shown in Table 8.2.
The frequency-compression scheme resulted in some perceptual
improvements. Certain fricative phonemes, such as /S/ and /j/, (shown in Table
8.2 in grey) were correctly identified by subjects more often with the frequency
compression device than with the conventional device. Unfortunately, the
experimental scheme also resulted in some reductions in perception. For
example, the phoneme /g/ was frequently mistaken for /z/. The recognition of /s/
was also reduced (shown in Table 8.2 in red). The last row of Table 8.2 shows
the number of times a phoneme was selected as a response. Some fricative
sounds such as /S/, /z/ and /v/ were selected more often as responses by
subjects when wearing the frequency-compression device. It is possible that
subjects were hearing more fricative-like sounds with the experimental scheme
but were unable to identify these correctly on every presentation.
Perhaps the experimental processing may have resulted in improved
intelligibility of speech with further training. Subjects’ experience with the
frequency-compression device was limited. Three of the subjects were not
hearing-aid users at the time of participation, and these subjects wore the
frequency-compression device infrequently. The logged time that they wore the
device was on average 2.2 hours per week. Based on the speech perception
- 145 results, it is possible that these subjects did not perceive a benefit when
Stimulus
wearing the device and chose to wear it irregularly.
CD response
p t
k
22
p 19
15 4 22
t
5 2 26
k
b
d
4
g
m
1
n
1
5
s
6
sh
1
z
7
f
3 20
ch
1
j
1 4
th
4
v
40 12 121
FrC response
p t
k
p 14 1 26
10 2 28
t
3 5 32
k
b
d
g
1
m
n
1 4
s
4
sh 1
z
3
f
ch 4 4 15
1
j
2 1
th
v
32 16 114
b
9
d
G
m
n
1
23 10
33 7
29
s
sh z
f
2
ch j
1
3
1
4
1
th
v
1
1
2
1
1
1
2
5
3
1
38 4
30 11
18 4
13 10 1
7
1 27
11 3
1
2 25
1
13 5
1 7
2 1
6
11 82 61 68 15 61 24 40
18 1
11 1
13
10 3
1
3
39 26 25
b
f
d
G
m
n
s
sh z
4
6
5
2
6
ch j
2
6
6
2 15
28 19
th
1
v
2
2
10 22 5
3 24 9
2 21
4
6
14
1
33 7
32 10
1
1
23
2
2
5 6
4 5
1
16 3
6
3
7
2
1 6
2 1 4
3
1
3
8 8
4
20 1
9 9
2 3 6
10
2 6
1
13
1 19
13 59 56 65 17 39 51 71 29 30 25 29 26
10
3
3
12
1
8
22
1
3
8
1
Table 8.2 Confusion matrices for the consonant test. Correct responses for each
phoneme are shown in bold type diagonally. The total number of times each phoneme
was selected as a response is shown in bold type at the bottom of each matrix.
- 146 – Chapter 8 Frequency compression outcomes
Subjectively, scores were higher in the APHAB questionnaire for the frequencycompression scheme for only one participant (S38). Unfortunately, the
subjective results do not seem to correlate with the objective speech test
results. For example, S38 reported preferring frequency compression, but
showed a score decrease in the binaural condition for both the word and
consonant test when wearing the device. In contrast, S37 indicated a strong
preference for the conventional device, yet showed some improvements in
speech perception for the binaural condition in quiet when wearing the
frequency-compression device.
The speech test results are consistent with those previously reported by
McDermott & Dean (2000) for six subjects with steeply-sloping losses who
showed no significant differences when frequency shifting was enabled
compared to when it was disabled. In that study, the tests included recognition
of monosyllabic words with proportional frequency shifting in which all
frequencies were lowered by a factor of 0.6.
The present results are inconsistent with those reported previously for an
identical signal-processing scheme (Simpson et al., 2005c). Differences in the
audiogram configurations across studies may account for the discrepancy. For
example, S35 had the best high-frequency hearing thresholds of the group and
was also the one subject who showed significant improvements when using the
frequency-compression device for tests in noise. Subjects who participated in
the Simpson et al. (2005c) study had, on average, some degree of lowfrequency hearing loss sloping to a severe-to-profound high-frequency loss for
frequencies above 2000 Hz. As noted, the current subjects had, on average,
near-normal hearing thresholds in the low frequencies, with a sharp drop in
thresholds to severe-profound levels for frequencies above 1000 Hz. It is likely
that these subjects had high-frequency dead regions based on the severity of
their hearing thresholds (Moore, 2001). Therefore, much of the compressed
signal used in this study may have been presented in the dead-region range of
hearing for these subjects. This may have caused part of the signal to be
inaudible, or audible but not helpful for discriminating the speech signals used in
the tests. The resulting lack of useful additional information may account for the
- 147 frequency-compression scheme showing no overall benefit for speech
perception in these experiments.
There may be a limit on how much information can be “squeezed” onto a
subject’s range of usable hearing. In Simpson et al. (2005c), the cut-off
frequencies were set between 1600 – 2500 Hz because this was close to the
lowest frequency at which the hearing thresholds were greater (worse) than 90
dB HL. In the current study, a lower cut-off frequency range (1250 - 1600 Hz)
was selected partly because of the different audiogram configurations of these
subjects. One possible effect of the steeper audiogram slopes is that some
components of the frequency-compressed signal were inaudible, and that cutoff frequencies below 1250 Hz could have resulted in an improvement in
speech perception. However, due to the nonlinearity of the compression, a
lower cut-off frequency would likely have resulted in a decrease in sound
quality. Any increase in high-frequency information might have been at the
expense of distorting lower-frequency signals, such as vowel sounds. As
described earlier, four of the subjects found the sound quality of the frequencycompression device unacceptable with a cut-off frequency of 1250 Hz. In these
cases, the cut-off frequency was increased to 1600 Hz.
Finally, the results of Experiment 2 may have been biased by subjects’
knowledge of Experiment 3. All subjects were informed of the possibility of
cochlear implantation prior to the frequency-compression trial. Subjects may not
been have favourable towards the frequency-compression scheme if they were
concerned that this may affect whether they were selected for Experiment 3.
However, only three subjects (S36, S37, S39) from Experiment 2 chose to go
ahead with implantation. The remaining four subjects showed results for
Experiment 2 which were consistent with subjects S36, S37, and S39. Namely,
that the frequency-compression scheme provided only minimal perceptual
changes for this group of 7 subjects.
Based on these observations, audiogram configuration is suggested to be an
important factor when recommending frequency compression. The frequency at
which the hearing loss becomes severe as well as the steepness of the slope of
the audiogram may be important. Listeners with a steeply-sloping hearing loss
can expect to receive limited benefit from the experimental frequency-
- 148 – Chapter 8 Frequency compression outcomes
compression scheme tested in Experiment 2. However, based on previously
reported results (Simpson et al., 2005c), some listeners with a more moderate
slope may obtain some additional cues about high-frequency signals when
wearing the frequency-compression device.
8.5.1 Conclusions
Experiment 1 (Chapter 7) found that while high levels of highfrequency audibility may provide speech benefits, this may be
at the expense of other practical factors, such as user comfort
and
feedback
oscillation.
Experiment
2
evaluated
the
performance of a frequency-compression device by comparing
the speech understanding abilities of seven hearing-impaired
listeners with steeply-sloping hearing losses in both quiet and
noisy conditions. The results can be summarized as follows.
1. Use of the frequency-compression scheme provided similar
recognition
of
monosyllabic
words
and
consonants
as
conventional hearing aids, on average.
2. Improvements when listening to sentences in noise were
shown
for
one
subject
when
wearing
the
frequency-
compression device.
3. The results of the APHAB questionnaire showed higher
global scores for four of the six subjects for the conventional
device over the frequency-compression scheme.
4. Results may have been impacted by the short time subjects
wore the device per week.
- 149 -
Chapter 9
Experiment 3
Combining electric and acoustic stimulation by means of a
frequency-place matched map
9.1
Introduction
Chapter 8 discussed how frequency compression did not prove to be effective
at providing high-frequency speech cues for a group of 7 listeners with steeplysloping losses. It was hoped that cochlear implantation may provide more
favourable results. Three subjects who participated in the frequencycompression trial opted to go ahead with cochlear implantation. A further two
subjects were recruited. This chapter describes the methods followed (section
9.2) and results found (section 9.3) for these 5 individuals who were implanted
with the Nucleus Freedom Contour Advance system. If useful hearing was
present post-operatively, subjects were fitted with an in-the-ear hearing aid
together with the speech processor in the implanted ear. All subjects continued
to wear a hearing aid in the contralateral ear. The acoustic frequency
corresponding to the most-apical electrode in terms of the pitch perceived was
determined with a pitch estimation task. The implant was programmed to
provide the listener with high-frequency information above the frequency at
which the acoustic hearing was no longer considered useful.
9.2
Methods and Procedures
An outline of the method followed for each subject is shown in Figure 9.1. Each
section of the outline is discussed in further detail below.
9.2.1 Subjects
Five hearing-impaired adults participated in the trial. Three subjects (S36, S37,
S39) had also participated in the frequency-compression study described in
Chapter 8. The remaining two subjects (S41, S42) were recruited from the
Melbourne Cochlear Implant Clinic. Relevant information about the subjects is
- 150 – Chapter 9 Combining electric and acoustic stimulation
provided in Table 8.I. Their pre-operative hearing threshold levels, measured
24 WEEKS POSTOPERATIVE
28 WEEKS
28 - 32
POSTWEEKS
POSTOPERATIVE
OPERATIVE
Pitch estimation training
Surgery: 2 week recovery
Testing:
1. Audiogram
2. Pitch estimation task
Fitting before start-up:
1. Place-matched MAP created.
2. Initial programming of both HAs
Fitting at start-up:
1. Implant test
2. Measurement of T- and C-levels
3. All 3 devices loudness-balanced
PLACE-MATCHED MAP
Testing:
1. Audiogram
2. Speech recognition: CNC words,
consonants, sentences in noise
3. APHAB questionaire
TIME
12 WEEKS POSTOPERATIVE
STARTUP
PRE-START-UP
PRESURGERY
conventionally under headphones, are shown in Figure 9.3.
CONVENTIONAL MAP
Testing:
1. Audiogram
2. Speech recognition: CNC words,
consonants, sentences in noise
PLACE-MATCHED MAP
Testing:
Speech recognition: CNC words,
sentences in noise
Subject given both MAPs
APHAB questionaire
Figure 9.1 Flow chart of the testing carried out for Experiment 3 for each of the 5 subjects
who participated in the study.
- 151 9.2.1.1 Criteria for recruitment
The criteria considered when approaching potential participants were developed
with the aim of recruiting subjects with the following characteristics:
1. Eighteen years of age or older.
2. Severe-to-profound sensorineural hearing loss for frequencies > 2000 Hz, but
with bilateral normal hearing thresholds, or a mild-to-moderate hearing loss, for
frequencies ≤ 1000 Hz. Thresholds for mid-frequencies would fall within the
mild-through-profound range.
3. Postlinguistic onset of bilateral severe-to-profound high-frequency
sensorineural hearing loss.
4. Lack of open-set word understanding, defined as aided (i.e., with
appropriately fitted hearing aids) CNC word recognition scores (mean of 2 lists)
between 0% and 30% when presented in quiet at 60-65 dBA in the ear to be
implanted.
5. Normal/patent cochlear anatomy.
6. English spoken as the primary language.
9.2.1.2 Ethics Approval
This study was approved by the Human Research and Ethics Committee at the
Royal Victorian Eye and Ear Hospital (Project No 99 / 362H). The test
procedure and aims of the study were discussed with each subject. Written
information regarding the study was given to each subject (Appendix A) and
informed consent was obtained before cochlear implantation was carried out.
Each subject was reimbursed for their travel costs by the Cooperative Research
Centre (CRC) for Cochlear Implant and Hearing Aid Innovation.
9.2.2 Surgical technique
A similar surgical technique 5 was followed as described by Lehnhart (1993). For
each subject, the round window membrane was identified by removing the lip of
5
The surgical technique followed as described by Lehnhart (1993) was modified by Mr Robert
Briggs. Surgery for all subjects was carried out by Mr Robert Briggs and Mr Stephen O’ Leary.
- 152 – Chapter 9 Combining electric and acoustic stimulation
the round window niche. The cochleostomy was sited inferior, rather than
anterior to the membrane. Bone was removed near the floor of the scala
tympani so that the spiral ligament and basilar membrane was not
compromised. Initially, a 1 mm diamond burr was used to mark the endosteum
of scala tympani, taking care to avoid opening the scala until drilling was
completed. Final bone removal was achieved with a 0.6 mm diamond burr.
Hyaluronic acid liquid gel (Healon) was used to clear bone dust during final
drilling and reduce perilymph leakage once the cochleostomy was opened.
Care was taken to avoid suctioning perilymph. The stimulator/receiver package
was positioned prior to completing the cochleostomy so that the electrode could
then be inserted and the cochleostomy sealed in as short a time as possible.
The Contour Advance electrode array was inserted using the “Advance off
Stylet” technique, which allows the electrode array to be inserted without
contacting the lateral wall of the cochlea or under surface of the basilar
membrane. This technique was designed by the manufacturer (Cochlear Ltd,
Lane Cove, N.S.W., Australia). To avoid over insertion, the electrode was
inserted approximately 17 mm with only the first of the three marker ribs inside
the cochleostomy. After inserting the array the cochleostomy was sealed with
fibrous tissue.
9.2.3 Pitch estimation testing
As shown in Figure 9.1, pitch-estimation testing was carried out pre- and postoperatively. The testing aimed to determine the acoustic frequency
corresponding to the most apical electrode (electrode 22) in terms of the pitch
perceived for each subject. Results were used to create a place-matched map.
For this map, the listener’s natural hearing provided low-frequency information,
and the implant supplemented this by providing only high-frequency information.
The procedure followed when carrying out the pitch-estimation testing is
described below.
An in-house software program 6 was used to carry out the testing. The output of
the computer soundcard was sent to the speech processor. Subjects heard the
6
The software program used to carry out pitch estimation testing was developed by Hugh
McDermott and Rodney Millard at the University of Melbourne.
- 153 acoustic stimuli via EARTone 5a insert earphones. Electric stimuli were
presented via the cochlear implant.
Test stimuli
The test stimuli consisted of pure tones with duration of 0.5 sec having
frequencies in the region of each subject’s acoustic hearing, and 5 electric
stimuli. The lower frequency boundary selected for testing was 125 Hz. The
upper frequency boundary was dependent upon the subject’s residual hearing.
Useful residual hearing was defined as a threshold of less than 90 dB HL for the
purposes of this study. Thus, the frequency assignment of the acoustic stimuli
ended at the frequency where acoustic hearing was no longer useful (i.e.,
thresholds ≥ 90 dB HL). The frequencies of the stimuli were spaced at thirdoctave intervals. The tones were ramped in level at each end with linear ramps
of duration 0.3 sec.
For the electric stimuli, electrodes most likely to correspond to acoustic low
frequencies were selected. For this reason, the most apical electrodes 22, 20,
18, 16, and 14 were selected initially. If the subject was unable to distinguish
between the pitch of these electrodes, then electrodes 22, 17, 12, 7, and 2 were
selected. The electric stimuli were steady pulse trains of duration 0.5 sec with
the same rate (900 Hz) as that of the subjects’ speech processor. Monopolar
mode (MP1+2) and minimum pulse width of 25 μs were used throughout the
testing procedure.
Test Procedure
A training run occurred before the subject was implanted to assist the subject in
becoming familiar with the task. Initially, all stimuli were loudness balanced to
ensure that the loudness of the stimuli did not affect each subject’s pitch ratings.
The highest-frequency acoustic stimulus was selected as the reference signal.
A loudness rating procedure was used (see Chapter 8 section 8.2.4.1 Fitting
method) to set the reference signal’s level. This signal was adjusted in level
until the subject perceived it to be “comfortably loud”. The subject was then
presented with two signals: the reference signal alternating with one of the test
signals. The subject heard these stimuli repeated continuously in quick
succession. The subject was instructed to adjust the loudness of the test
- 154 – Chapter 9 Combining electric and acoustic stimulation
stimulus by means of a toggle switch until the two stimuli were equal in
loudness. The subject was required to “bracket” the equal-loudness level by
setting the level of the test stimulus first too high and then too low before
converging on a final value.
For each test stimulus, initially two separate loudness balances were obtained,
one with the test stimulus pre-set to a level slightly too loud, and the other with it
pre-set slightly too soft. If the difference between these two loudness-balanced
levels was less than 3 dB (or ± 3 levels), then the average of these two levels
was used as the final level for that test stimulus. If the difference was greater,
another two loudness balances were obtained as above, and the final level for
the test stimulus was set to the average of the four values.
To ensure that subjects used an appropriately wide number range when giving
pitch estimates, the subject was played the reference tone and told that this
tone had a pitch corresponding to the number 50. The subject was then
instructed to give each tone they heard a number in relation to the reference
tone. For example, a pitch perceived as twice as high as the reference signal
should be given a pitch estimate close to 100. Subjects were instructed that
they could give pitch estimates of greater than 100 or use fractions if necessary.
The stimuli were repeated 10 times each, and presented to the subject in a
random order. In addition, the levels of the stimuli were randomized by the
software. The total range of level variation was ± 1 dB (or ± 1 level) around the
level the subject had selected as “comfortably loud”. As it was unknown how the
surgical procedure would impact on each subject’s hearing thresholds, the
entire procedure was repeated separately for each ear. If for example, surgery
resulted in a large hearing loss in the implanted ear, pitch estimation could still
be carried out in the contralateral ear for mapping purposes.
Following surgical implantation of the device and an adequate healing period,
the implant was activated (usually 2 to 3 weeks after surgery). Threshold (T)
and Comfort (C) levels were measured for electrical stimulation on each
electrode at a rate of 900 Hz. Monopolar mode (MP1+2) and minimum pulse
width of 25 μs were selected. The ACE strategy (see Chapter 5, section 5.4.1
- 155 The Advanced Combination Encoder (ACE) speech-processing strategy) was
selected with 8 maxima.
Pitch estimation testing was carried out before the subject had any listening
experience via the speech processor of the cochlear implant. Initially, all stimuli
were loudness balanced as described for the pre-operative training. An identical
procedure to the one described above was carried out. The subject was played
the acoustic reference tone and told that this tone had a pitch corresponding to
the number 50. The subject was instructed to give each tone they heard a
number in relation to the reference tone. The stimuli were repeated 10 times
and presented to the subject in a random order. If the subject had useful
residual hearing in the implanted ear, the acoustic stimuli were presented to the
implanted ear. If the subject had no useful residual hearing in the implanted ear,
the acoustic stimuli were presented to the contralateral ear.
The results of the pitch estimation task were used to create individual maps for
each subject. The fitting procedure followed for each subject is described below.
9.2.4 Pre-operative fitting procedure
9.2.4.1 Frequency allocations and level adjustment
Subjects who participated in the study had up to three separate listening
devices which required fitting and adjustment. These were the cochlear implant
(CI), the hearing aid on the implanted side (HAi), and the hearing aid on the
contralateral side (HAc). The following section describes how each of these
devices was programmed.
Contralateral hearing aid
Each subject was fitted with the Phonak Supero 412 as described in Chapter 8
(see section 8.2.3.1 Fitting method) prior to implantation. No changes were
made to this fitting.
Ipsilateral hearing aid
Subjects were fitted with an in-the-ear (ITE) hearing aid (Phonak Savia 33 FS) if
hearing thresholds were aidable in the implanted ear. For this study, aidable
hearing was defined as two or more hearing thresholds in the implanted ear of
less than 90 dB HL at the following frequencies: 125, 250, 500, 750, 1000,
- 156 – Chapter 9 Combining electric and acoustic stimulation
1500, 2000, 3000, 4000, 6000, and 8000Hz. The cut-off point for acoustic
hearing was the frequency which was closest to either equaling 90 dB HL, or
falling just below 90 dB HL.
A fitting was created using NAL-NL1 (Byrne et al., 2001) as the suggested fitting
target. The gain for frequencies at which the subject had no useful residual
hearing was set to the minimum possible value as determined by constraints of
the software. The amplitude compression was left at values determined by the
software. Features such as feedback and noise cancellation were turned off.
This fitting was saved into the device before the subject arrived for the initial
fitting appointment.
Cochlear implant
The current clinical default map in the Melbourne Cochlear Implant Clinic is
based on the ACE processing strategy, with a rate of 900 Hz, and 8 maxima.
The default frequency-to-electrode allocations for this map are shown in Table
9.1.
Two maps were created with different frequency-to-electrode allocations. The
first map, known as the place-matched map, aimed to assign frequency
channels to the electrode array so as to supplement the subject’s acoustic
sensitivity as well as mimicking the tonotopic organization of the ear. The
subject wore this map in weeks 1-12 post-operatively. The second map, known
as the conventional map, was assigned the default clinical frequency-toelectrode allocations as shown in Table 9.1. The subject wore this map in
weeks 12-24 post-operatively. It was expected that subjects’ results on the
perceptual tests would be affected by the amount of time they were exposed to
the sound of the implant. To compensate for learning effects of testing as well
as acclimatization effects from wearing the implant, subjects were switched
back to the place-matched map from weeks 24-28 post-operatively.
Before creating the place-matched map the following tests were carried out
post-operatively:
1. An audiogram was carried out in the implanted ear. The following frequencies
were tested: 125, 250, 500, 750, 1000, 1500, 2000, 3000, 4000, 6000, and
8000 Hz.
- 157 2. Pitch estimation testing. The results were plotted as pitch estimate (linear
scale) versus frequency (log scale). A regression line was fitted to the acoustic
data using the following equation:
y = y 0 +(a * ln ( x ))
Equation 6
Where y corresponds to the vertical axis of the graph, x to the horizontal axis,
y 0 corresponds to the vertical offset of the fitted line, a is the gradient of the
line, and ln is the natural logarithm.
The acoustic frequency corresponding to electrode 22 in terms of the pitch
perceived was estimated by means of the following formula:
f =e
P − y0
a
Equation 7
where f is the approximate acoustic frequency, e is 2.718, P is the mean pitch
estimate value, y 0 is the y-axis intercept, and a is the gradient of the plotted
line.
The assignment of acoustic frequencies for all three devices for all subjects for
the place-matched map is shown in Figure 9.2. The hearing aid in the nonimplanted ear was left unchanged for all subjects. An ITE hearing aid amplified
frequencies in the implanted ear with useful residual hearing as defined above.
Only two subjects (S37, S41) were fitted with an ITE hearing aid as well as the
implant in the implanted ear. For S37, frequencies between 125 – 750 Hz in the
implanted ear were amplified with an ITE hearing aid. For S41, frequencies
between 125 – 500 Hz in the implanted ear were amplified with an ITE hearing
aid.
- 158 – Chapter 9 Combining electric and acoustic stimulation
Electrode
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
Lower
Frequency
Boundary
(Hz)
188
313
438
563
688
813
938
1063
1188
1313
1563
1813
2063
2313
2688
3063
3563
4063
4688
5313
6063
6938
Upper
Frequency
Boundary
(Hz)
313
438
563
688
813
938
1063
1188
1313
1563
1813
2063
2313
2688
3063
3563
4063
4688
5313
6063
6938
7938
Centre
Frequency
(Hz)
250.5
375.5
500.5
625.5
750.5
875.5
1000.5
1125.5
1250.5
1438
1688
1938
2188
2500.5
2875.5
3313
3813
4375.5
5000.5
5688
6500.5
7438
Bandwidth
(Hz)
125
125
125
125
125
125
125
125
125
250
250
250
250
375
375
500
500
625
625
750
875
1000
Table 9.1 Frequency-to-electrode allocations in the current clinical default map, used in
the Melbourne Cochlear Implant Clinic, assuming all 22 electrodes are active.
For the CI, the acoustic frequency of electrode 22 was calculated as described
above. The frequency range of the electrodes was assigned to agree closely
with the calculated acoustic frequency of electrode 22. Electrode 22 was
assigned to the centre frequency in the current default filter bank design (shown
in Table 9.1) that was closest to the calculated acoustic frequency of electrode
22. The highest frequency filter-band (7438 Hz), as shown in Table 9.1, was
always assigned to the most basal electrode (electrode 1) for all subjects.
Respectively, electrodes 22-1 were assigned a range with centre frequencies of
625.5 – 7438 Hz for S36, 625.5 – 7438 Hz for S39, 500.5 – 7438 Hz for S41,
and 875.5 – 7438 Hz for S42.
- 159 -
Hearing aid: Non-implanted ear
S36
Hearing aid: Non-implanted ear
Hearing aid:
Implanted ear
E22 = acoustic frequency of 630 Hz
S37
Hearing aid: Non-implanted ear
S39
Hearing aid: Non-implanted ear
Hearing aid:
Implanted ear
S41
Hearing aid: Non-implanted ear
S42
0.125
0.5
1
2
4
Acoustic Input Frequency (kHz)
8
Figure 9.2 Schematic representation of the frequency allocations for the CI and both
hearing aids for the place-matched map for the 5 subjects who participated in
Experiment 3.
- 160 – Chapter 9 Combining electric and acoustic stimulation
Electrodes were disabled if the calculated acoustic frequency fell within the
subject’s useful residual hearing range. This was the case for one subject
(S37). For this subject, electrode 22 was disabled. Electrodes 21-1 were
assigned to a range of 750.5 – 7438 Hz.
All subjects were changed to the conventional map at 12 weeks postoperatively. When creating the conventional map, the aim was to resemble the
current clinical default map in Melbourne (ACE processing strategy, with a rate
of 900 Hz, and 8 maxima) as closely as possible. For this reason, a frequency
range with the default frequency-to-electrode allocations shown in Table 9.1
was selected (250.5 – 7438 Hz). When fitting, other than T-and C-levels, no
other parameters were changed for either map. No changes were made to the
frequency assignment of any hearing aid.
9.2.5 Post-operative fitting procedure
At the start-up appointment, the transmitting coil was fitted to the subject and an
appropriate magnet strength was determined.
The subject was given the following instructions:
“Today we will be fitting your cochlear implant. The aim of today’s session is to
find a setting that you find comfortable. To achieve this, we will be playing you a
variety of sounds before we switch the implant on. We expect that the implant
may sound strange when we first switch it on.”
The subject was then asked to turn off any hearing device they may have been
wearing.
T and C levels were measured for each electrode using monopolar mode at a
rate of 900 Hz.
For T-levels, the clinician aimed to measure the lowest stimulus level where a
response always occurred. A method similar to the standard clinical method
proposed by Carhart & Jerger (1959) was followed. The current level for each
electrode was increased until a hearing sensation was reported by the subject.
Once a response was found, the level was adjusted in descending steps of 10
current levels (CLs) and ascending steps of 5 CLs. Several responses were
recorded and averaged to determine the final T-level for each electrode.
- 161 When measuring C-levels, subjects were given the categorical list of 9 loudness
levels which was described in Chapter 8 (section 8.2.4.1 Fitting method).
Starting from T-level, the current level was adjusted upwards in steps of 5 CLs
until the subject reported the sound to be “Loud but O.K.”
To ensure that sound through the implant was comfortable and audible, it was
necessary to check that maximum hearing sensation across electrodes were
perceived as equally loud by the subject. Once T -and C-levels had been
determined for all electrodes, C-levels were played to subjects in groups of
three electrodes to ensure they were similar in loudness across all electrodes.
The standard clinical method was followed where the subject was played the Clevels on three of the most-apical adjacent electrodes. The following group of
three included one electrode from the previous group. This was repeated until
each electrode had been included in the measure. Any electrode which was
perceived to vary in loudness when compared to the group was adjusted. The
level of the electrode reported to vary was adjusted in steps of 3 current levels
(CL) up or down depending on whether the subject reported the stimuli to be
softer or louder than the adjacent electrodes. C-levels for the group of three
electrodes were then played again to the subject. This procedure was repeated
until all electrodes were perceived as equally loud.
As most stimulation occurs somewhere between T- and C-level, loudness
growth was also assessed by sweeping across electrodes at an intensity level
halfway between T- and C-levels. For the current study, this was selected to be
50% of the dynamic range. An identical procedure to the one described above
for C-levels was carried out.
Before switching the implant on, it was anticipated that subjects may experience
loudness summation effects as electrodes would be stimulated together, rather
than one at a time. To initially compensate for the effects of loudness
summation, T- and C-levels were reduced by 20% of the dynamic range.
The subject was asked to turn on all hearing aid/s they were wearing. The
speech processor was then turned on, and the subject asked to provide
feedback on the sound quality of the device.
- 162 – Chapter 9 Combining electric and acoustic stimulation
Adjustments were made to the map based on the subject’s feedback. If the
subject reported loudness discomfort, then T- and C-levels were dropped in
steps of 5% of the dynamic range.
9.2.5.1 Loudness balancing: Measuring a comfortable loudness level for
speech
Once the subject was comfortable with the initial sound quality of the device,
loudness balancing procedures were carried out with all three devices. The
general aim of the fitting was to ensure that speech was both comfortable and
audible for subjects. It was assumed that this would be best achieved by
ensuring that speech at conversational levels was perceived as equally loud
through each device. The following procedure was followed.
Test stimuli
The material consisted of male-and-female speaker continuous speech
obtained from the AH “Speech and Noise for Hearing Aid Evaluation” CD
(National Acoustic Laboratories., 2000). It was played at 60 dBA, a level
corresponding to between conversational and raised speech (Skinner et al.,
1997), measured at a distance of 1 m from the loudspeaker.
Test procedure
The sensitivity of the implant was left at the recommended setting. The subject
was asked to turn off any other device they may have been wearing. The
subject was asked to indicate on the loudness scale a level for the speech
material. Both T -and C-levels were adjusted together in steps of 5% of the
dynamic range until the subject indicated a level that was “comfortable but
slightly soft” on the loudness scale. It was hoped that each subject would
perceive speech to be a “comfortable” level. A level softer than this, i.e.
“comfortable but slightly soft” was chosen as it was anticipated that there would
be loudness summation effects when all three devices were turned on together.
The implant speech processor was then turned off. The contralateral hearing
aid (HAc) was turned on. Subjects were asked to give an estimate of how loud
the speech material was through the hearing aid by using the loudness scale.
No changes were made to the HAc if the subject reported speech to be
“comfortable but slightly soft”. If the subject perceived speech to be at a level
- 163 different from this, the overall gain of the hearing aid was adjusted in steps of 3
dB until the subject reported speech to be “comfortable but slightly soft”.
The HAc was then turned off, and the ipsilateral hearing aid (HAi) switched on.
Subjects were again asked to give an estimate of how loud the speech material
was through the hearing aid by using the loudness scale. No changes were
made to the HAi if the subject reported speech to be “comfortable but slightly
soft”. If the subject perceived speech to be at a level different from this, the
overall gain of the hearing aid was adjusted in steps of 3 dB until the subject
reported speech to be “comfortable but slightly soft”.
9.2.5.2 Loudness balancing: Measuring maximum loudness level
An additional aim of the fitting procedure was to ensure that external sounds
would not be perceived as uncomfortably loud for subjects. The procedure
followed is described below.
Test stimuli
The material consisted of ICRA noise sampled from the Australian Hearing
“Speech and Noise for Hearing Aid Evaluation” CD. ICRA noise is 3-band
Schröder filtered English speech, which is unintelligible and has idealized
spectra according to ANSI (ANSI, 1993). A description of the ICRA noise can be
found in Dreschler et al. (2001). The noise was looped and played at 80 dBA
measured at a distance of 1 m from the loudspeaker.
Test procedure
The sensitivity of the implant was left at the recommended setting. Using the
loudness scale, if the subject reported the noise to be “uncomfortably loud”, or
above, C-levels were turned down in steps of 5% of the dynamic range until the
subject reported the sound to be “Loud but OK”. The implant was then turned
off, and the HAc turned on. The same procedure used to check maximum
loudness levels for the implant was carried out with each HA separately. The
Maximum Power Output (MPO) of each hearing aid was adjusted in steps of 3
dB until the subject reported the noise to be “Loud but OK”. The procedure was
repeated for the HAi. Finally, both hearing aids as well as the CI were turned on
to check for loudness summation effects. The maximum loudness level was
checked again using the above stimulus and procedure.
- 164 – Chapter 9 Combining electric and acoustic stimulation
The subject was advised to use all three devices as much as possible and to
experiment with the volume controls of the HA/s and sensitivity control of the CI
at home to find a setting that was most comfortable for listening to speech in
quiet and noisy conditions. Each subject was seen for weekly visits after the
initial switch-on session. No changes were made to the CI or the HA/s if the
subject reported no complaints about the sound quality of any device. T- and Clevels were checked, but not changed unless there was a large difference
(greater than 5 % of the dynamic range on average) across sessions.
9.2.6 Monitoring of hearing thresholds
An audiogram was carried out for each ear at 2, 12, and 24 weeks postoperatively. Air conduction thresholds were obtained at the following
frequencies: 125, 250, 500, 1000, 1500, 2000, 3000, 4000, 6000, and 8000 Hz.
9.2.7 Speech recognition testing
Chapter 8 (see section 8.2.3.2 Speech test stimuli and methods) described the
speech stimuli and methods used for this study. They were a test of
monosyllabic CNC words and consonants in quiet, as well as sentence testing
in noise. Speech recognition testing was carried out at 12, 24, and 28 weeks
post-operatively. Due to time constraints, consonant testing was not repeated at
28 weeks post-operatively. Table 9.2 shows the tests carried out and conditions
included for each of these time points.
9.2.8 Subjective assessment
Each subject was given the ABHAB questionnaire (Cox et al., 1995) to
complete at 12 weeks post-operatively. The subject was asked to compare their
listening experience with hearing aids pre-operatively to the cochlear implant
together with hearing aid/s post-operatively. It was also of interest to determine
whether subjects perceived sound quality differences between the two maps as
well as which map was preferred overall. Therefore, once speech perception
testing was completed at 28 weeks post-operatively, each subject was given
both maps as well as the APHAB questionnaire. The full-range map was
assigned to Program 1, and the place-matched map to Program 2. Subjects
were not told about the map assignment order. They were instructed to listen to
- 165 both programs in a variety of listening situations before completing the
questionnaire.
Time interval
Test condition
CNC
Words
Sentences
in noise
Consonants
Pre-operatively
(Conventional device)
Both HAs alone
Left HA alone
Right HA alone
Both HAs alone
CI alone
CIHAi
CIHAc
CIHA/s
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
12 weeks post-operatively
(Place-matched map)
24 weeks post-operatively
(Conventional map)
28 weeks post-operatively
(Place-matched map)
28 -32 weeks postoperatively
Both HAs alone
CI alone
CIHAi
CIHAc
CIHA/s
CIHA/s
CI alone
Subject given
both MAPs
APHAB
*
*
Table 9.2 Speech recognition and subjective test conditions included for the 5 subjects
who participated in the study.
9.3
Results
The study design of the current experiment included multiple speech perception
measurements within subjects over time. Table 9.2 shows all the conditions
which were included in the study. Statistical analyses resulted in a large number
of comparisons between these measurements. Those comparisons thought to
be most relevant to the aims and objectives of the study will be discussed
below. However, not all conditions could be measured for all subjects. The
reason for this was the partial or total loss of useful hearing in the implanted ear
for some subjects post-operatively. Statistical interpretation in some cases was
therefore limited to conditions where data was available for all subjects.
- 166 – Chapter 9 Combining electric and acoustic stimulation
9.3.1 Pre- and post-operative audiograms
Pre-and post-operative hearing thresholds are shown for the 5 subjects in
Figure 9.3. For all subjects, the average hearing thresholds in the frequency
range of 125 – 1000 Hz was calculated for the measurement taken at 24 weeks
post-operatively. This value was subtracted from the average pre-operative
hearing thresholds in the implanted ear for the same frequency range. A value
of 5 dB was added for hearing thresholds for which no response was measured
at the audiometer’s maximum output (i.e., if the maximum output of the
audiometer was 110 dB, then 115 dB was the value entered when calculating
the average). The final value was reported as the average hearing loss for each
subject.
On average, at 24 weeks post-operatively, S36 lost 52 dB, S37 lost 16 dB, S39
lost 34 dB, S41 lost 31 dB, and S42 lost 28 dB. Hearing thresholds for the
implanted ear remained stable over time for S36, S37, and S39. For these
subjects, thresholds did not shift more than 10 dB at any frequency. S41
showed a worsening in hearing thresholds over time. An average loss of 11 dB
was measured for S41 at 2 weeks post-operatively, compared to 31 dB at 24
weeks post-operatively. S42 showed an improvement in thresholds over time.
An average loss of 43 dB was measured for S42 at 2 weeks post-operatively,
compared to 28 dB at 24 weeks post-operatively.
- 167 0
Non-implanted ear
Implanted ear pre-operative
Implanted ear 2 wks post-operative
Implanted ear 12 wks post-operative
Implanted ear 24 wks post-operative
Hearing level (dB HL)
20
40
60
80
100
120
0
S36
S41
S37
S39
Hearing level (dB HL)
20
40
60
80
100
120
0
0.125
Hearing level (dB HL)
20
0.25
0.5 0.75 1
1.5 2
3 4
6 8
Frequency (kHz)
40
60
80
100
120
S42
0.125
0.25
0.5 0.75 1
1.5 2
3 4
6 8
Frequency (kHz)
Figure 9.3 Pre- and post-operative audiogram results for the 5 subjects who participated
in the study. For each subject, lines in blue indicate the left ear, and lines in red indicate
the right ear. The mean measurement for the non-implanted ear and pre-operative
thresholds for the implanted ear are shown by unfilled circles connected by solid lines.
Unfilled triangles connected by dotted lines show thresholds for the implanted ear
measured at 2 weeks post-operatively. Unfilled squares connected by dashed lines show
thresholds for the implanted ear measured at 12 weeks post-operatively. Filled circles
connected by solid lines show thresholds for the implanted ear measured at 24 weeks
post-operatively. Error bars indicate one standard deviation. Arrows indicate levels
which were limited by the maximum output of the audiometer.
- 168 – Chapter 9 Combining electric and acoustic stimulation
9.3.2 Pitch estimation results
Pitch estimation was carried out to determine the acoustic frequency
corresponding to the most-apical electrode in terms of the pitch perceived.
Results from the pitch estimation task were used to create individual maps for
each subject as described in section 9.2.4.1 (Cochlear implant). The task
consisted of initially loudness-balancing acoustic and electric stimuli (see
section 9.2.3 Pitch estimation testing). The subject was then instructed to give
each tone they heard a number in relation to the reference tone. As both S37
and S41 had usable residual hearing in the implanted ear, both electric and
acoustic stimuli were presented to the same ear. For the remaining 3 subjects
(S36, S39, S42) acoustic and electric stimuli were presented in opposite ears.
The most-apical electrodes 22, 20, 18, 16, and 14 were selected for all subjects,
with the exception of S41. For this subject the results obtained with these
electric stimuli were difficult to interpret in terms of pitch. Instead, electrodes 22,
17, 12, 7, and 2 were selected and used in a subsequent test.
The pitch estimation results carried out post-implantation are shown for the 5
subjects in Figure 9.4. For each subject, pitch estimates obtained for acoustic
stimuli are shown on the left, while pitch estimates obtained for electric stimuli
are shown on the right. In general, pitch estimates for both acoustic and electric
stimuli increased monotonically for each subject, with lower pitch estimates
corresponding to more apical electrodes and lower frequencies. Linear
regression analysis found this to be statistically significant with p values of less
than 0.002 for all subjects.
The regression line fitted to the data for the acoustic stimuli was used to
calculate the pitch-matched frequency corresponding to electrode 22 for all
subjects as described in section 9.2.4.1 (Cochlear implant). Visual inspection of
S41’s data revealed that a straight line would be a good fit to all points except
for the stimulus of 160 Hz. S41 provided a higher pitch estimate for 160 Hz then
for 200 Hz, although these two estimates were not significantly different. For
this reason, the regression line was only fit to the higher estimates in order to
extrapolate beyond the highest acoustic frequency in order to get a closer
correspondence with the electric pitch perception of electrode 22. The pitchmatched frequency corresponding to electrode 22 was estimated to equal 579
- 169 Hz for S36, 680 Hz for S37, 610 Hz for S39, 601 Hz for S41, and 887 Hz for
S42.
100
80
S36
60
40
20
400
100
80
R
L
0
500
630
800
1000
22
E22 = 579 H z
20
18
16
14
12
S37
60
Pitch Estimate
40
20
L
0
315
100
80
400
500
630
L
E22 = 680 Hz
800
22
20
R
L
1000
22
18
16
14
12
S39
60
40
20
0
400
100
80
500
630
800
E22 = 610 Hz
20
18
16
14
S41
60
40
20
160
100
80
R
R
0
200
250
315
400
22
E22 = 601Hz
17
12
7
2
S42
60
40
20
R
L
E22 = 887Hz
0
400
500
630
800
F re q u e n cy (H z)
1000
22
20
18
16
14
E le c tro d e
Figure 9.4 Pitch estimation results measured at 2 weeks post-operatively for the 5
subjects who participated in Experiment 3. Pitch estimates obtained for acoustic stimuli
are shown on the left, while pitch estimates obtained for electric stimuli are shown on the
right. For each subject, the ear selected to present the acoustic and electric stimuli is
indicated by the symbol “R” for the right ear, and “L” for the left ear. The regression line
used to calculate the pitch-matched frequency corresponding to electrode 22 is shown
by the solid black line for the acoustic stimuli.
- 170 – Chapter 9 Combining electric and acoustic stimulation
9.3.3 Speech perception results
For the sake of brevity, the following abbreviations have been used when
describing the speech perception results. The abbreviation HA/s refers to
speech perception testing which was conducted with the subject wearing only
hearing aids. For this condition post-operatively, only two of the subjects (S37,
S41) wore binaural hearing aids. The remaining three subjects (S36, S39, S42)
wore only one hearing aid in the non-implanted ear. The abbreviation CI alone
refers to speech perception testing which was conducted with the subject
wearing only their cochlear implant. The abbreviation CIHA/s refers to speech
perception testing which was conducted with the subject wearing their cochlear
implant together with their hearing aid or aids. Subjects S37 and S41 wore
binaural hearing aids whereas the remaining three subjects wore just one
hearing aid in the non-implanted ear.
For CNC word testing in quiet, additional conditions of the cochlear implant
together with the contralateral hearing aid (CIHAc), and the cochlear implant
together with the ipsilateral hearing aid (CIHAi) were included for the two
subjects (S37, S41) who wore a hearing aid in the implanted ear. These
conditions were included for these two subjects to determine whether there
were any perceptual differences in combining the CI signal with the acoustic
hearing in the implanted ear, combining the CI signal with the acoustic hearing
in the non-implanted ear, and combining the CI signal with the acoustic hearing
from both ears.
- 171 9.3.3.1 CNC words in quiet
Figure 9.5 shows the percent of all phonemes (vowels and consonants) in the
CNC word test recognized correctly by each subject for each condition. Figures
9.6, 9.7, and 9.8 show the percent correctly recognized of consonants,
consonants containing frication (i.e., /f/, /s/, /S/, /v/, /z/, /T/, /tS/, and /Z/), and
vowels, respectively, extracted from the same data. (In the following, the above
subset of consonants will be referred to, for brevity, as fricatives, although the
last two are usually classified as affricates. Fricatives were included the analysis
as they provide an indication of subjects’ performance on speech parts
containing mostly high-frequency information).
Post-operatively, all subjects showed an improvement in phoneme scores when
compared to pre-operative results. On average, subjects scored 44% (SD 9%)
pre-operatively. This increased to 66% (SD 23%) with the CI alone at 12 weeks
post-operatively, 77% (SD 8%) at 24 weeks post-operatively, and 69% (SD
12%) at 28 weeks post-operatively. Scores in the CIHA/s condition were
generally higher than for CI alone. Mean scores for the group with CIHA/s were
77% (SD 16%) at 12 weeks post-operatively, 82% (SD 7%) at 24 weeks postoperatively, and 83% (SD 10%) at 28 weeks post-operatively.
A two-way ANOVA was carried out on subjects’ raw phoneme, consonant,
fricative, and vowel scores. A significant interaction was found between the
scheme and the subject factors (p < 0.001). Each subject’s data was analyzed
separately with pair-wise comparisons using the Holm-Sidak method. Results
for phonemes, consonants, fricatives, and vowels can be seen in Tables 9.3,
9.4, 9.5, and 9.6 respectively.
- 172 – Chapter 9 Combining electric and acoustic stimulation
Across the group, pre-operative phoneme, consonant, and vowel scores were
significantly higher (p < 0.05) than HA/s alone scores at 12 and 24 weeks postoperatively. Pre-operative fricative scores did not differ significantly (p ≥ 0.05)
from HA/s alone scores at 12 and 24 weeks post-operatively. Pre-operative
phoneme, consonant, fricative and vowel scores were found to be significantly
lower than the CI alone, and CIHA/s conditions at 12, 24, and 28 weeks postoperatively.
Subjects generally performed best when wearing the CI together with their
hearing aid/s. For phonemes and vowels, CIHA/s scores were significantly
higher than CI alone scores at 12, 24, and 28 weeks post-operatively. For
consonants CIHA/s scores were significantly higher than CI alone scores at 12
and 28 weeks (place-matched map) post-operatively. At 24 weeks
(conventional map) post-operatively, CIHA/s scores were not significantly
different to CI alone scores. For fricatives, there was no significant difference
between CIHA/s scores and CI alone scores at 12 weeks (place-matched map)
and 24 weeks (conventional map) post-operatively. At 28 weeks (placematched map) post-operatively, CIHA/s scores were significantly higher than CI
alone scores.
Generally, mean scores with the conventional map were higher when subjects
were wearing the CI alone. For phonemes, consonants, fricatives, and vowels,
CI alone scores at 12 weeks post-operatively (place-matched map) did not differ
significantly from CI alone scores at 28 weeks (place-matched map) postoperatively. For phonemes, consonants, fricatives, and vowels, CI alone scores
at 12 weeks (place-matched map) post-operatively were significantly lower than
CI alone scores at 24 weeks (conventional map) post-operatively.
- 173 For phonemes and vowels, CI alone scores at 24 weeks (conventional map)
post-operatively were significantly better than CI alone scores at 28 weeks
(place-matched map) post-operatively. Consonant and fricative CI alone scores
at 24 weeks (conventional map) post-operatively did not differ significantly from
CI alone scores at 28 weeks (place-matched map) post-operatively.
Scores for the conventional and place-matched map did not appear to be
significantly different when subjects were wearing the CI together with hearing
aid/s. For phonemes, consonants, and fricatives, CIHA/s scores at 12 weeks
post-operatively (place-matched map) did not differ significantly from CIHA/s
scores at 24 weeks (conventional map) post-operatively. Vowel scores in the
CIHA/s condition were significantly higher at 24 weeks (conventional map) postoperatively when compared with CIHA/s scores at 12 weeks post-operatively
(place-matched map). For phonemes, consonants and vowels, CIHA/s scores
at 12 weeks (place-matched map) post-operatively were significantly lower than
CIHA/s scores at 28 weeks (place-matched map) post-operatively. There was
no significant difference between fricative scores in the CIHA/s condition at 12
and 28 weeks post-operatively. For phonemes, consonants, fricatives, and
vowels, CIHA/s scores at 24 weeks (conventional map) post-operatively were
not significantly different from CIHA/s scores at 28 weeks (place-matched map)
post-operatively. For phonemes, consonants, fricatives, and vowels, there was
no significant difference in scores between the conditions of CIHAc, CIHAi, and
CIHA/s for the two subjects (S37, S41) who wore a hearing aid in the implanted
ear.
- 174 – Chapter 9 Combining electric and acoustic stimulation
100
HA/s alone
CI alone
CIHAi
CIHAc
CIHA/s
Phonemes correctly recognized (%)
90
80
70
60
50
40
30
20
10
S36
S37
S39
S41
S42
All Ss
0
100
Phonemes correctly recognized (%)
90
80
70
60
50
40
30
20
10
Phonemes correctly recognized (%)
0
100
90
80
70
60
50
40
30
20
10
0
Pre-op
12
24
28 wks post-op
Pre-op
12
24
28 wks post-op
Figure 9.5 Phoneme scores for the CNC word test for the 5 subjects who participated in
Experiment 3. Scores for the HA/s alone condition are shown by the unfilled columns.
Scores with the CI alone are shown by the filled green columns. Scores for the CIHA/s
condition are shown by the filled purple columns. For those two subjects (S37, S41) who
wore hearing aids binaurally post-operatively scores in the CIHAi condition are shown by
the unfilled diagonally-lined columns. Scores in the CIHAc condition are shown by the
filled grey columns. The grey shaded area indicates scores at 12 and 28 weeks postoperatively when subjects were listening with the place-matched map. Scores averaged
across subjects are shown in the lower right panel. Error bars indicate one standard
deviation.
- 175 12 WEEKS
HA/S ALONE
CI ALONE
CIHA/s
HA/S ALONE
CI ALONE
CIHA/s
PLACE-MATCHED MAP
28 WEEKS
CIHA/s
CI ALONE
PLACE-MATCHED MAP
CONVENTIONAL MAP
24 WEEKS
12 WEEKS
PREOPERATIVE
SCORE
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
*
*
*
*
*
*
*
*
*
*
*
*
*
POST-OPERATIVE SCORE
24 WEEKS
PLACE-MATCHED MAP
HA/S
CI
ALONE ALONE CIHA/S
CI ALONE
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
CONVENTIONAL MAP
HA/S
CI
ALONE ALONE CIHA/S
28 WEEKS
PLACEMATCHED
MAP
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
Table 9.3 ANOVA results for phonemes scores in the CNC word test for the 5 subjects
who participated in Experiment 3. Asterisk symbols indicate statistical significance (p <
0.05).
- 176 – Chapter 9 Combining electric and acoustic stimulation
Consonants correctly recognized (%)
100
HA/s alone
CI alone
CIHAi
CIHAc
CIHA/s
90
80
70
60
50
40
30
20
10
S36
S37
S39
S41
S42
All Ss
0
Consonants correctly recognized (%)
100
90
80
70
60
50
40
30
20
10
Consonants correctly recognized (%)
0
100
90
80
70
60
50
40
30
20
10
0
Pre-op
12
24
28 wks post-op
Pre-op
12
24
28 wks post-op
Figure 9.6 Consonant scores for the CNC word test for the 5 subjects who participated in
Experiment 3. Scores for the HA/s alone condition are shown by the unfilled columns.
Scores with the CI alone are shown by the filled green columns. Scores for the CIHA/s
condition are shown by the filled purple columns. For those two subjects (S37, S41) who
wore hearing aids binaurally post-operatively scores in the CIHAi condition are shown by
the unfilled diagonally-lined columns. Scores in the CIHAc condition are shown by the
filled grey columns. The grey shaded area indicates scores at 12 and 28 weeks postoperatively when subjects were listening with the place-matched map. Scores averaged
across subjects are shown in the lower right panel. Error bars indicate one standard
deviation.
- 177 12 WEEKS
HA/S ALONE
CI ALONE
CIHA/s
HA/S ALONE
CI ALONE
CIHA/s
PLACE-MATCHED MAP
28 WEEKS
CIHA/s
CI ALONE
PLACE-MATCHED MAP
CONVENTIONAL MAP
24 WEEKS
12 WEEKS
PREOPERATIVE
SCORE
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
POST-OPERATIVE SCORE
24 WEEKS
PLACE-MATCHED MAP
HA/S
CI
ALONE ALONE CIHA/S
CONVENTIONAL MAP
HA/S
CI
ALONE ALONE CIHA/S
28 WEEKS
PLACEMATCHED
MAP
CI ALONE
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
Table 9.4 ANOVA results for consonant scores in the CNC word test for the 5 subjects
who participated in Experiment 3. Asterisk symbols indicate statistical significance (p <
0.05).
- 178 – Chapter 9 Combining electric and acoustic stimulation
Fricatives correctly recognized (%)
100
90
80
70
60
50
40
30
20
10
0
Fricatives correctly recognized (%)
100
90
80
70
60
50
40
30
20
10
0
Fricatives correctly recognized (%)
100
90
80
70
60
50
40
30
20
10
0
HA/s alone
CI alone
CIHAi
CIHAc
CIHA/s
S36
S37
S39
S41
S42
All Ss
Pre-op
12
24
28 wks post-op
Pre-op
12
24
28 wks post-op
Figure 9.7 Fricative scores for the CNC word test for the 5 subjects who participated in
Experiment 3. Scores for the HA/s alone condition are shown by the unfilled columns.
Scores with the CI alone are shown by the filled green columns. Scores for the CIHA/s
condition are shown by the filled purple columns. For those two subjects (S37, S41) who
wore hearing aids binaurally post-operatively scores in the CIHAi condition are shown by
the unfilled diagonally-lined columns. Scores in the CIHAc condition are shown by the
filled grey columns. The grey shaded area indicates scores at 12 and 28 weeks postoperatively when subjects were listening with the place-matched map. Scores averaged
across subjects are shown in the lower right panel. Error bars indicate one standard
deviation.
- 179 -
12 WEEKS
HA/S ALONE
CI ALONE
CIHA/s
HA/S ALONE
CI ALONE
CIHA/s
PLACE-MATCHED MAP
28 WEEKS
CIHA/s
CI ALONE
PLACE-MATCHED MAP
CONVENTIONAL MAP
24 WEEKS
12 WEEKS
PREOPERATIVE
SCORE
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
POST-OPERATIVE SCORE
24 WEEKS
PLACE-MATCHED MAP
HA/S
CI
ALONE ALONE CIHA/S
CONVENTIONAL MAP
HA/S
CI
ALONE ALONE CIHA/S
28 WEEKS
PLACEMATCHED
MAP
CI ALONE
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
Table 9.5 ANOVA results for fricative scores in the CNC word test for the 5 subjects who
participated in Experiment 3. Asterisk symbols indicate statistical significance (p < 0.05).
- 180 – Chapter 9 Combining electric and acoustic stimulation
Vowels correctly recognized (%)
100
90
80
70
60
50
40
30
20
10
0
Vowels correctly recognized (%)
100
90
80
70
60
50
40
30
20
10
0
Vowels correctly recognized (%)
100
90
80
70
60
50
40
30
20
10
0
HA/s alone
CI alone
CIHAi
CIHAc
CIHA/s
S36
S37
S39
S41
S42
All Ss
Pre-op
12
24
28 wks post-op
Pre-op
12
24
28 wks post-op
Figure 9.8 Vowel scores for the CNC word test for the 5 subjects who participated in
Experiment 3. Scores for the HA/s alone condition are shown by the unfilled columns.
Scores with the CI alone are shown by the filled green columns. Scores for the CIHA/s
condition are shown by the filled purple columns. For those two subjects (S37, S41) who
wore hearing aids binaurally post-operatively scores in the CIHAi condition are shown by
the unfilled diagonally-lined columns. Scores in the CIHAc condition are shown by the
filled grey columns. The grey shaded area indicates scores at 12 and 28 weeks postoperatively when subjects were listening with the place-matched map. Scores averaged
across subjects are shown in the lower right panel. Error bars indicate one standard
deviation.
- 181 12 WEEKS
HA/S ALONE
CI ALONE
CIHA/s
HA/S ALONE
CI ALONE
CIHA/s
PLACE-MATCHED MAP
28 WEEKS
CIHA/s
CI ALONE
PLACE-MATCHED MAP
CONVENTIONAL MAP
24 WEEKS
12 WEEKS
PREOPERATIVE
SCORE
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
*
*
*
*
*
*
POST-OPERATIVE SCORE
24 WEEKS
PLACE-MATCHED MAP
HA/S
CI
ALONE ALONE CIHA/S
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
CONVENTIONAL MAP
HA/S
CI
ALONE ALONE CIHA/S
28 WEEKS
PLACEMATCHED
MAP
CI ALONE
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
Table 9.6 ANOVA results for vowel scores in the CNC word test for the 5 subjects who
participated in Experiment 3. Asterisk symbols indicate statistical significance (p < 0.05).
- 182 – Chapter 9 Combining electric and acoustic stimulation
9.3.3.2 Consonants in quiet
Results for the consonant test are shown in Figure 9.9. A two-way ANOVA was
carried out on subjects’ raw scores. Statistical significance for the various factor
combinations are shown in Table 9.7. Each subject’s data was analyzed
separately with pair-wise comparisons using the Holm-Sidak method as a
significant interaction was found between the scheme and the subject factors (p
< 0.001). Across the group, pre-operative scores were not found to differ
significantly from the HA/s alone score measured at 12 and 24 weeks postoperatively. Visual inspection shows a large improvement in scores for all
subjects when pre-operative scores are compared with CI alone and CIHA/s
scores at both 12 and 24 weeks post-operatively. This score increase was
found to be significant (p < 0.05) for all subjects at both 12 and 24 weeks postoperatively. Across the group, there was a large difference between scores in
the HA/s alone condition compared with the CI alone and CIHA/s conditions.
This too was found to be significant at 12 and 24 weeks post-operatively. No
significant differences were found between the CI alone and CIHA/s conditions
at either 12 or 24 weeks post-operatively. Across the group, there was a trend
for CI alone scores measured at 12 weeks post-operatively to be lower than CI
alone scores measured at 24 weeks post-operatively. However, CIHA/s scores
measured at 12 weeks were generally higher than CIHA/s scores measured at
24 weeks post-operatively. Neither of these trends were shown to be significant.
- 183 -
Pre-Op
Consonants correctly recognized (%)
CI alone
100
HA/s alone
CIHA/s
90
80
70
60
50
40
30
20
10
0
S36
S37
S39
S41
S42
Mean
Figure 9.9 Results for the consonant test for the 5 subjects who participated in
Experiment 3. Pre-operative scores with binaural hearing aids are shown by the unfilled
striped columns. Scores with hearing aid/s (HA/s) alone post-operatively are shown by
the unfilled columns. Scores with the cochlear implant (CI) alone are shown by the filled
green columns. Scores with the hearing aid/s together with the cochlear implant (CIHA/s)
are shown by the filled purple columns. The grey shaded area indicates scores at 12
weeks post-operatively when subjects were listening with the place-matched map.
Scores averaged across subjects are shown in the rightmost columns, with error bars
indicating one standard deviation.
- 184 – Chapter 9 Combining electric and acoustic stimulation
POST-OPERATIVE SCORE
12 WEEKS
CI ALONE
CI ALONE
CIHA/s
CONVENTIONAL MAP
24 WEEKS
HA/S ALONE
CIHA/s
PLACE-MATCHED MAP
12 WEEKS
HA/S ALONE
PREOPERATIVE
SCORE
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
PLACE-MATCHED MAP
HA/S
ALONE
CI
ALONE
CIHA/S
*
*
*
*
*
*
*
*
*
*
*
*
24 WEEKS
CONVENTIONAL
MAP
HA/S
ALONE
CI
ALONE
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
Table 9.7 ANOVA results for the consonant test for the 5 subjects who participated in
Experiment 3. Asterisk symbols indicate statistical significance (p < 0.05).
- 185 -
9.3.3.3 Sentences in noise
The signal-to-noise ratio (SNR) in dB obtained for each of the subjects for the
conditions of HA/s alone, CI alone, and CIHA/s measured at 12, 24, and 28
weeks post-operatively are shown in Figure 9.10. As described in Chapter 8
(see section 8.2.3.2.3 Sentence recognition in noise), sentences were
presented combined with noise. An adaptive procedure was carried out to
determine the SNR for a target score of 50% words correct. Subjects required a
minimum score of 75% in quiet for the test to be carried out. One of the subjects
(S42) obtained a score of less than 75% in quiet pre-operatively. At 12 weeks
post-operatively, her score in quiet had improved to greater than 75%.
Therefore, testing in noise was carried out with this subject for the first time at
12 weeks post-operatively. Testing in noise was carried out with subject S39
pre-operatively. However, this subject showed a decrease in sentence scores in
quiet at 12 weeks post-operatively. As she scored less than 75% in quiet,
further testing in noise was not carried out. At 24 weeks post-operatively, S39’s
understanding in quiet had improved to allow for further testing in noise.
Subjects’ understanding of sentences was dependent on the device/s they were
wearing during testing. The test was easier if all devices were worn and more
difficult if a device was removed. For this reason, an SNR could not be obtained
for three subjects (S39, S41, S42) post-operatively when they were wearing
HA/s alone and for one subject (S39) when wearing the CI alone. This implies
that in order for the subject to obtain a score of 50% correct in noise, the level of
the noise had to be reduced to the point of inaudibility. An SNR value would be
meaningless at such low noise levels. Arrows on Figure 9.10 represent when an
SNR could not be obtained.
A two-way ANOVA was carried out on subjects’ raw scores. For those
conditions where testing was not carried out or an SNR could not be obtained, a
dummy SNR value of 35 dB was entered. The value was selected based on the
audibility function (W) in SII calculations, where the dynamic range of speech is
assumed to equal 30 dB (Pavlovic et al., 1984). A level outside of the 30 dB
range is assumed not to contribute to intelligibility. An assumption was made
that noise levels at an SNR of 35 dB were largely inaudible to subjects. Results
- 186 – Chapter 9 Combining electric and acoustic stimulation
can be seen in Table 9.8. Each subject’s data was analyzed separately with
pair-wise comparisons using the Holm-Sidak method as a significant interaction
was found between the scheme and the subject factors (p < 0.001).
Pre-operative scores for the group were found to be significantly better (p <
0.05) than post-operative HA/s alone scores at 12 and 24 weeks. HA/s alone
scores at 12 weeks post-operatively were not found to be significantly different
to HA/s alone scores at 24 weeks post-operatively. For the group, an
improvement in SNR was observed post-operatively. CI alone scores measured
at 24 weeks post-operatively were significantly better than pre-operative scores.
There was no significant difference between pre-operative scores and CI alone
scores at 12 and 28 weeks post-operatively. The best SNR result was found in
the CIHA/s condition. Pre-operative scores were significantly worse than CIHA/s
scores measured at 12, 24, and 28 weeks post-operatively. In addition, the
CIHA/s condition was significantly better than CI alone for 12, 24, and 28 weeks
post-operatively.
Similarly to the results reported for CNC words in quiet, some differences were
seen between the place-matched and conventional maps in the CI alone
condition. As a group, subjects performed better with the conventional map than
the place-matched map when wearing the CI alone. No significant differences
were found between CI alone scores at 12 (place-matched map) and 28 (placematched map) weeks post-operatively. CI alone scores at 24 (conventional
map) weeks post-operatively were significantly better than CI alone scores
measured at 12 (place-matched map) and 28 (place-matched map) weeks postoperatively. However, the addition of a hearing aid revealed no difference in
scores between the two maps. No significant differences were found between
CIHA/s scores at 12 (place-matched map) and 24 (conventional map) weeks
post-operatively, 12 (place-matched map) and 28 (place-matched map) weeks
post-operatively, and 24 (conventional map) and 28 (place-matched map)
weeks post-operatively.
Pre-op
12 wks
Post-op
24 wks
28 wks
- 187 -
HA/s alone
CI alone
CIHA/s
S37
12 wks
24 wks
Post-op
28 wks
S36
Pre-op
Did not test
S41
Pre-op
12 wks
24 wks
Post-op
28 wks
S39
Did not test
All Ss
S42
0
5
10
15
20
25
SNR (dB)
30
35
40
45
5
10
15
20
25
30
35
40
45
SNR (dB)
Figure 9.10 Mean signal-to-noise ratios (SNRs) obtained by the 5 subjects who
participated in Experiment 3. Unfilled columns show SNRs obtained with hearing aids
(HA/s) alone, filled green columns show SNRs obtained with the cochlear implant (CI)
alone, and filled purple columns show SNRs obtained with cochlear implant together
with hearing aid/s (CIHA/s). Arrows indicate where an SNR could not be obtained
because subjects did not obtain a 50% correct sentence score at low noise levels.
- 188 – Chapter 9 Combining electric and acoustic stimulation
POST-OPERATIVE
CI ALONE
CI ALONE
CI ALONE
CIHA/s
PLACE-MATCHED MAP
28 WEEKS
CIHA/s
CONVENTIONAL MAP
24 WEEKS
HA/S ALONE
CIHA/s
PLACE-MATCHED MAP
12 WEEKS
HA/S ALONE
PREOPERATIVE
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
S36
S37
S39
S41
S42
Mean
12 WEEKS
24 WEEKS
PLACE-MATCHED MAP
HA/S
CI
ALONE ALONE CIHA/S
CONVENTIONAL MAP
HA/S
CI
ALONE ALONE CIHA/S
28 WEEKS
PLACEMATCHED
MAP
CI ALONE
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
Table 9.8 ANOVA results for the sentence in noise test for the 5 subjects who
participated in Experiment 3. Asterisk symbols indicate statistical significance (p < 0.05).
- 189 -
9.3.4 Subjective assessment
The APHAB questionnaire was scored in an identical manner to that described
in section 8.4.4 (Subjective assessment). At 12 weeks post-operatively subjects
were asked to compare their impressions of their listening abilities pre- and
post-operatively. For each subscale, total scores from subjects’ impressions
post-operatively were subtracted from the total score obtained from subjects’
impressions pre-operatively. These values are shown in Figure 9.11 for each of
the 5 subjects. A positive score indicates a preference for post-operative
listening conditions whereas a negative score indicates a preference for preoperative listening conditions. All subjects showed preferences for postoperative listening conditions in the subscales of ease of communication and
global scores. Preferences for post-operative listening conditions were also
shown by four subjects (S36, S37, S39, S41) in the subscale of background
noise, three subjects (S36, S37, S41) in the subscale of reverberation and two
subjects (S36, S42) for the subscale of aversion. A preference for pre-operative
listening conditions was reported by three subjects (S37, S39, S41) for the
subscale aversion. A two-way ANOVA run on subjects’ average scores found a
statistically significant difference between pre- and post-operative listening
conditions for the subscales of ease of communication (df = 1, p = 0.006) and
background noise (df = 1, p = 0.03). No significant differences were found for
the subscales of reverberation (df = 1, p = 0.07), aversion (df = 1, p = 0.47) and
global (df = 1, p = 0.27).
At 28 weeks post-operatively, both the place-matched and conventional maps
were saved in the subject’s speech processor. Subjects were asked to compare
their impressions of their listening abilities with “Program 1” (conventional map)
and “Program 2” (place-matched map). For each subscale, total scores of
subjects’ impressions with the conventional map were subtracted from the total
score obtained from subjects’ impressions with the place-matched map. These
values are shown in Figure 9.12 for each of the 5 subjects. A positive score
indicates a preference for the conventional map, whereas a negative score
indicates a preference for the place-matched map.
- 190 – Chapter 9 Combining electric and acoustic stimulation
Post-operative
80
60
40
20
Pre-operative
0
-20
-40
-60
EC
BN
RV
AV
Global
-80
S36
S37
S39
S41
S42
Figure 9.11 Preference scores from the APHAB questionnaire at 12 weeks postoperatively provided by the 5 subjects who participated in the study. For each subject,
the subscale Ease of Communication (EC) is shown by the filled black columns, the
subscale Background Noise (BG) is shown by the filled pink columns, the subscale
Reverberation (RV) is shown by the unfilled columns, the subscale Aversion (AV) is
shown by the filled green columns, and Global scores are shown by the filled grey
columns. As shown on the vertical axis, negative values indicate a preference for preoperative listening conditions, whereas positive values indicate a preference for postoperative listening conditions.
Overall, preference ratings were small. S37 showed a preference for the placematched map in the subscale of background noise, and for the conventional
map in the subscale of reverberation. S41 indicated a slight preference for the
place-matched map in the subscale of aversion. S39 showed a preference for
the conventional map in the subscales of background noise, reverberation,
aversion, and global, but for the place-matched map in the subscale of ease of
communication. S36 and S42 showed no preference for either map in any
subscale. Global scores showed no preference for either map for four (S36,
S37, S41, S42) of the subjects. A two-way ANOVA run on subjects’ average
scores found no significant differences across the group between the
conventional and place-matched maps for all subscales (Ease of
communication [df = 1, p = 0.53], background noise [df = 1, p = 0.83],
- 191 reverberation [df = 1, p = 0.24], aversion [df = 1, p = 0.98], global [df = 1, p =
Place-matched map
Conventional map
0.37]).
15
10
5
EC
BN
RV
AV
Global
0
-5
-10
S36
S37
S39
S41
S42
Figure 9.12 Preference scores from the APHAB questionnaire carried out after 28 weeks
post-operatively provided by the 5 subjects who participated in Experiment 3. Subjects
were given both the conventional map and the place-matched map and asked to compare
them. For each subject, the subscale Ease of Communication (EC) is shown by the filled
black columns, the subscale Background Noise (BG) is shown by the filled pink columns,
the subscale Reverberation (RV) is shown by the unfilled columns, the subscale
Aversion (AV) is shown by the filled green columns, and Global scores are shown by the
filled grey columns. As shown on the vertical axis, negative values indicate a preference
for the place-matched map, whereas positive values indicate a preference for the
conventional map.
9.4
Discussion
The experiment presented above aimed to investigate the following three
questions. Firstly, can hearing be preserved in listeners with steeply-sloping
losses who choose to undergo cochlear implantation? Secondly, how does
cochlear implantation affect speech perception for this group of listeners?
Thirdly, would the use of a place-matched map provide an advantage for these
listeners above a conventional full frequency-range map?
In answer to the first question, four of the five subjects (S37, S39, S41, S42)
who participated in this study did have some measurable hearing thresholds
post-operatively. On average, these four subjects lost 27 dB in the frequency
- 192 – Chapter 9 Combining electric and acoustic stimulation
range of 125 – 1000 Hz. The remaining subject (S36) lost all hearing. Two of
the subjects (S37, S41) were fitted with an ITE hearing aid post-operatively as
both subjects had two or more hearing thresholds of less than 90 dB HL in the
implanted ear. Hearing thresholds for the implanted ear remained stable over
time for three of the subjects (S36, S37, and S39). For these subjects,
thresholds did not shift more than 10 dB at any frequency. Of the remaining two
subjects, one subject (S41) showed deterioration in hearing thresholds over
time and one subject (S42) showed an improvement in thresholds over time.
The reasons for the drop in S41’s threshold over time are unclear. Little is
known at this stage about the long-term effects of electrode insertion for hearing
preservation. One possibility is that the loss of hearing could have been
attributed to an undetected bacterial or viral infection.
Studies reporting hearing threshold changes over time with cochlear
implantation were reviewed in Chapter 6. The results of Experiment 3 are most
comparable with those reported by James et al. (2005). The Nucleus Contour
Advance electrode array used in that study has the same electrode design as
the Freedom device selected for Experiment 3. A similar surgical technique was
also followed. In that study, a low-frequency drop in hearing of 23-33 dB was
reported for a group of 12 subjects. Two subjects lost all hearing. The results of
Experiment 3 are not as favourable as those reported for the short-array Hybrid
device (Gantz et al., 2003; Gantz et al., 2004; Gantz et al., 2005) or the Med-El
device with a limited insertion depth (Gstoettner et al., 2004; Kiefer et al., 2005).
An average loss of 9 dB was reported for the Hybrid device on a group of 24
subjects, whereas studies with the Med-El device reported most subjects had a
loss of between 10-20 dB. If all the current literature (Gantz et al., 2003; Gantz
et al., 2004; Gantz et al., 2005; Gstoettner et al., 2004; James et al., 2005;
Kiefer et al., 2005; Skarzynski et al., 2003; von Ilberg et al., 1999) on the
subject of preserving residual hearing with steeply-sloping losses is grouped
together with the current study, a total of 77 patients have been implanted. Nine
(12%) of these patients experienced a total loss of hearing.
A more important question is how the preservation of residual hearing affects
patient outcomes. If the advantage of the addition of low-frequency hearing in
the implanted ear were known for certain, the risks to the patient would be more
- 193 clearly understood. On the other hand, if the preservation of low-frequency
hearing added very little benefit for the patient, the goal of hearing preservation
would not be as crucial.
In the current study, all five subjects showed improvements in speech
perception at 12, 24, and 28 weeks post-operatively in quiet in tests of word and
consonant recognition. When compared to pre-operative scores, for the word
test, the group’s average phoneme score increased by 22 percentage points
with CI alone and by 33 percentage points with CIHA/s at 12 weeks postoperatively. For the consonant test, the group’s mean scores improved by 40
percentage points with CI alone, and 47 percentage points with CIHA/s at 12
weeks post-operatively.
Loss of hearing did not appear to be a predictor for the subjects’ outcomes.
S36, for example, lost all hearing in the implanted ear, yet showed large
improvements in speech perception. This patient obtained a phoneme score
with the CI alone of 86% correct at 12 weeks post-operatively, compared to a
score of 50% pre-operatively. However, loss of hearing did impact on word
understanding for the group when subjects wore their HA/s alone postoperatively. For the word test, across the group, pre-operative phoneme,
consonant, and vowel scores were significantly better than post-operative HA/s
alone scores at 12 and 24 weeks post-operatively. The exception was fricative
scores, which showed no significant differences between pre-operative and
post-operative scores at 12 or 24 weeks. However, fricative scores were low for
the group pre-operatively and in the HA/s condition. The non-significant result
may be a reflection of scores nearing chance levels. As a group, subjects did
not perceive many fricative sounds correctly pre-operatively, as well as when
wearing HA/s alone post-operatively. HA/s alone scores for the consonant test
appeared unaffected by implantation as pre-operative scores were not found to
differ significantly from HA/s alone scores at 12 and 24 weeks post-operatively.
Larger subject numbers would be required to determine whether shorter and/or
thinner arrays with a reduced insertion depth offer a greater chance at hearing
preservation. However, no matter which array is selected, total or partial hearing
loss remains a risk for the patient. As shown above, even partial loss of hearing
is likely to affect speech perception, assuming the patient chooses to wear only
- 194 – Chapter 9 Combining electric and acoustic stimulation
hearing aids. Future patients with similar or more low-frequency hearing would
need to be fully informed of this risk.
For testing in quiet, a bimodal benefit was found for the word test, but not for the
consonant test. At 28 weeks post-operatively, across the group, subjects
obtained a significantly higher score with CIHA/s when compared to CI alone for
word testing in quiet for phonemes, consonants, fricatives and vowels. For the
consonant test, there was no significant difference between the CI alone and
CIHA/s condition at 12 and 24 weeks post-operatively. For the two subjects
(S37, S41) who were fitted with an ITE hearing aid in the implanted ear, scores
in the CIHAi, CIHAc, and CIHA/s conditions for the word test were not
significantly different. It appears that, for this test at least, these two subjects did
not show any advantage with three devices in quiet compared to the remaining
subjects who were fitted with just two devices. A previous comparison of two
versus three devices has been reported by Gantz et al. (2005). In that study,
long-term speech results were shown for 8 patients. A reproduction of these
speech results is shown in Figure 9.13.
The conditions of CI alone, CIHAi, and CIHA/s are plotted against pre-operative
HA/s alone scores for a word test in quiet. Statistical significance was not
reported in the study. However, visual inspection found 5 of the subjects
obtained their highest score in the CIHA/s condition and 3 subjects obtained
their highest score in the CIHAi condition. The difference in scores between the
two conditions (CIHAi versus CIHA/s) appeared to be large for 4 of the subjects
with 3 of these obtaining their highest scores in the CIHA/s condition. As stated
earlier, it is unknown whether there was a significant difference in scores
between these two conditions and the CI alone. Due to the small number of
subjects, it is not possible to be conclusive about whether the addition of an ITE
is beneficial for speech perception in quiet. However, it seems reasonable to
assume from the current result and the Gantz et al. (2005) study that three
devices (CI combined with two hearing aids) does not necessary result in a
higher speech score than two devices (CI combined with one hearing aid), at
least for quiet conditions.
- 195 -
Figure 9.13 Monosyllabic CNC word scores for Hybrid patients. Pre-operative scores for
the implanted ear are shown by the solid black bar, pre-operative scores with binaural
HAs are shown by the stippled bar, scores with CI alone are shown by the black bars
with white dots, scores in the CIHAi condition are shown by the bars with horizontal
lines, and scores in the CIHA/s condition are shown by the bars with diagonal lines.
(Figure reproduced from Gantz et al. 2005).
Chapter 6 described previous studies (Gantz et al., 2005; James et al., 2005;
Kiefer et al., 2005) which have reported group speech perception scores for
subjects with steeply-sloping losses who have undergone implantation with the
aim of preserving residual hearing. The results together with those of
Experiment 3 are summarized in Table 9.9. For the current study, word scores
(i.e. not phoneme scores) at 24 weeks post-operatively are shown. It should be
noted that many of the studies did not report speech scores. The data shown in
Table 9.9 were derived from visual inspection of figures shown for each of the
above studies, and therefore the percentages quoted may not be exact. Some
caution should be used when comparing the current study’s findings to
previously reported results. All studies used a small number of subjects which
makes generalizations inappropriate. Speech test material differed in each
study dependent on the language spoken by the subjects. Two of the studies
(Gantz et al., 2005; Kiefer et al., 2005) reported speech results after subjects
had worn the implant for a longer time than the current study. It is likely that
subjects’ scores would improve over time as they adjust to the sound of the
implant (Hollow et al., 2006). This, together with the other factors mentioned
above makes direct comparison between the studies unfeasible.
- 196 – Chapter 9 Combining electric and acoustic stimulation
James et al.
(2005)
Nucleus Contour
Advance
Mono- and
disyllabic words
Gantz et al.
(2005)
Nucleus
Hybrid
Monosyllabic
words
Kiefer et al.
(2005)
MED-EL
COMBI 40+
Monosyllabic
words
English (Australian
accent)
German
(monosyllabic)
French and
Spanish
(disyllabic)
English
(American
accent)
German
6 months
3 months
12 months
5
6
Long term (6,
9, or 12
months)
8
16%
14%
25%
7%
53%
63% (2 Ss only)
40%
44% (3 Ss only)
42%
64%
CIHAc
71% (2 Ss only)
Not measured
Not measured
CIHA/s
Preoperative
CI alone
63%
7 - 23%
Not measured
5 - 30%
73%
17 - 43%
56%
62%
Not
measured
Not
measured
0 - 35%
2 - 85%
35 - 63% (3 Ss
only)
18 - 75%
25 - 83%
CIHAi
35-70%
59 – 67% (2 Ss
only)
30 - 85%
CIHAc
64 - 77% (2 Ss only)
Not measured
Not measured
CIHA/s
50 – 77%
Not measured
67 - 90%
30 - 90%
Not
measured
Not
measured
Study
Electrode array
selected
Speech test
selected
Current
Nucleus Freedom
Language
speech test
performed in
Length of CI
experience
(months)
Number of
subjects (Ss)
Mean Preoperative
word
score CI alone
CIHAi
Word
score
range
Monosyllabic words
13
Table 9.9 Summary of word score results in quiet for the current and previously reported
studies.
In general though, subject outcomes appear broadly similar across all studies.
As shown in Table 9.9, mean scores as well as the range of scores were largely
similar across test conditions despite differences in electrode array and time of
measurement. When comparing pre-operative scores to CI alone, the large
majority of subjects obtained an improvement in word scores. However, 1
subject in the current study, 2 subjects in the Gantz et al. (2005) study, and 1
subject in the James et al. (2005) study showed only small differences in
scores. In addition, one subject reported by James et al. (2005) obtained a
reduction in scores with the CI alone. When the CI was combined with a HA in
either the ipsi- or contralateral ear, all subjects obtained an increase in word
scores when compared to pre-operative scores. The use of a HA together with
- 197 a CI appears to be beneficial for the large majority of patients. Patient scores
were either similar or improved in the CIHAi, CIHAc, or CIHA/s conditions when
compared to CI alone scores across all studies. Only one subject in the Kiefer
et al. (2005) study obtained higher scores for CI alone than for CIHAi.
Unfortunately, no previously reported study included all test conditions of CI
alone, CIHAi, CIHAc, and CIHA/s. Gantz et al. (2005) did not report the
combination of CIHAc, and Kiefer et al. (2005) and James et al. (2005) did not
report the conditions of CIHAc or CIHA/s. It is therefore not possible to
determine which combination of HA device and implant was best for subjects
and whether the addition of an HA in the implanted ear offers any advantage
above that of wearing a HA in the contralateral ear only.
In summary, testing in quiet for Experiment 3 found large perceptual
improvements post-operatively when compared to pre-operative scores for the
group. Generally, subjects perform similarly or better when wearing their CI
together with hearing aid/s when compared to CI alone scores. Results appear
to be comparable to the findings of previously reported studies. The lowfrequency acoustic hearing provides most patients with an advantage. However,
this advantage could come from either the implanted or non-implanted ear.
Whilst one could assume that “the more hearing the better”, there is currently
limited evidence to prove that a CI patient with residual low-frequency hearing in
both ears is at more of an advantage than a CI patient with residual lowfrequency hearing in one ear.
Analysis of test scores in noise found a similar result to the word tests in quiet.
Pre-operative scores were significantly better than HA alone scores measured
at 12, 24, and 28 weeks post-operatively. As for testing in quiet, it is likely that
subjects’ loss of low-frequency hearing would have affected their scores. Across
the group, CI alone scores and CIHA/s scores at 12 and 24 weeks postoperatively were significantly better than pre-operative scores. The best result
was obtained when the measurement was carried out with each subject wearing
their HA/s together with the CI. Across the group, CIHA/s scores were
significantly higher than CI alone scores at 12, 24, and 28 weeks postoperatively. Chapter 6 summarizes results of previous studies where testing of
CI recipients was carried out in noise. Again, there are some limitations when
- 198 – Chapter 9 Combining electric and acoustic stimulation
comparing these results to previous studies (Gantz et al., 2005; James et al.,
2005; Kiefer et al., 2005) due to differences in test stimuli and methods. The
results, together with those of with the current study are summarized in Table
9.10. Results at 24 weeks post-operatively are shown for the current study.
Of the 6 subjects tested by James et al. (2005), 5 obtained an improved SNR
when CI alone scores were compared to pre-operative scores. When CI alone
scores were compared to CIHAi scores for the 3 subjects who were fitted with a
HA in the implanted ear, all 3 obtained better scores in the CIHAi condition. One
subject performed worse in noise compared to their pre-operative score for both
the CI alone and CIHAi condition. Mean sentence-correct scores in noise for a
group of 13 patients were reported to improve from 8% pre-operative to 31%
with CI alone (Kiefer et al., 2005). The best result was found in the CIHAi
condition with a mean score of 74%. Results reported by Gantz et al. (2005) do
not include pre-operative and CI alone scores. Rather, SNRs for a group of 8
Hybrid electrode recipients when wearing CIHA/s were compared to those of a
group of 20 conventional-electrode CI recipients. The Hybrid group obtained a
significantly better result in noise when compared to the conventional CI group.
Whilst this is a favourable result, it is unknown how this score compared to preoperative and CI alone data for the same group of subjects. In addition, it is
unknown whether the conventional CI recipients had useful hearing in their
contralateral ear.
Most “traditional” CI users perform poorly in background noise (Friesen et al.,
2001; Fu et al., 1998; Nelson et al., 2003). This is thought to be due in part to
the poor spectral resolution a CI offers in comparison to normal hearing. As
many processing schemes rely on speech envelope detection, much of the fine
temporal and spectral resolution of the signal is lost to the user. All results
described above show that most patients perform best in noise when a HA is
used together with their CI. It is likely that the acoustic hearing via the HA is
able to provide the patient with additional spectral and temporal cues that are
important for speech understanding, especially in noise, even though the
acoustic hearing is limited to low-frequency information only. The result
supports findings in which the addition of acoustic low-frequency hearing may
provide better pitch perception compared to hearing with a CI alone, and this is
- 199 expected to provide benefits in speech perception in backgrounds of competing
talkers as well as for music perception (Gantz et al., 2005; Turner et al., 2004).
Study
Electrode array selected
Speech test stimuli
Current
Nucleus
Freedom
Speech:CUNYlike sentences
Noise: 8-talker
babble
James et al.
(2005)
Nucleus Contour
Advance
Speech:Sentences
Noise: Not
described
Gantz et al.
(2005)
Nucleus Hybrid
Speech:Spondee
words
Noise: 2-talker
babble
Kiefer et al.
(2005)
MED-EL
COMBI 40+
Speech:HSM
sentences
Noise:
Unknown
Speech test method
SNR at 50%
correct
recognition
SNR at 50%
correct recognition
SNR at 50%
correct
recognition
Score obtained
with SNR of 10
dB
Language speech test
performed in
English
(Australian
accent)
6 months
German, French
and Spanish
German
5
6
English
(American
accent)
Long term (6, 9,
or 12 months)
8
13
14 dB
16 dB
Not measured
8%
31 dB
Not measured
Not measured
Not measured
CI alone
12 dB
6 dB
Not measured
CIHAi
Not measured
5 dB (3 Ss only)
Not measured
31%
74% (8 Ss
only)
CIHAc
Not measured
Not measured
Not measured
Not measured
CIHA/s
Preoperative
9 dB
Not measured
-2 dB
Not measured
13 – 40 dB
13 - 20 dB
Not measured
0 - 35%
HA alone
17 – 40 dB
Not measured
Not measured
Not measured
CI alone
9 – 21 dB
Not measured
CIHAi
Not measured
-1 - 9 dB
-2 - 18 dB (3 Ss
only)
Not measured
0 - 60%
39 - 90 % (8 Ss
only)
CIHAc
Not measured
Not measured
Not measured
Not measured
CIHA/s
6 – 16 dB
Not measured
Not measured
Not measured
Length of CI experience
(months)
Number of subjects (Ss)
Mean SNR (dB) Preor sentences
operative
correct (%)
HA alone
SNR score
range (dB) or
sentences
correct (%)
3 months
12 months
Table 9.10 Summary of speech testing in noise for the current and previously reported
studies.
Subjectively, APHAB results at 12 weeks post-operatively found that all patients
experienced less difficulties in the subscales of “ease of communication”,
- 200 – Chapter 9 Combining electric and acoustic stimulation
“background noise”, and “reverberation” when comparing their listening
experience pre- and post-operatively. This result is consistent with the large
improvement in speech perception for all subjects. The only subscale in which
patients indicated a preference for pre-operative listening conditions was
“aversion”. This is not surprising. Although the CI has provided improved
communication, it has also resulted in subjects experiencing more external
noise, not all of which may be desirable.
The third question of interest in the experiment described above was whether
the use of a place-matched map provided any advantage for these listeners
above that of a conventional full frequency-range map. Many past studies
(described in Chapter 6) have found evidence that the tonotopic shift in
frequency that is presumed to occur with CI processing could result in a
decrease in speech recognition (Baskent et al., 2003; Baskent et al., 2004;
Baskent et al., 2005; Dorman et al., 1997; Friesen et al., 2001; Fu et al., 1999).
Many of these studies were simulations with normally-hearing listeners or
carried out through the CI alone. It was thought that CI processing which
mimicked the normal ear in terms of frequency-place allocations could result in
the best outcome for patients in terms of speech perception as well as the time
taken to adjust to the sound quality of the implant. In the present study, for word
testing in quiet and sentence testing in noise, CI alone scores with the
conventional map were significantly better than CI alone scores with the placematched map. Considering the place-matched map provided subjects with a
narrower frequency range, and consequently less speech information, than the
conventional full-range map, it is not surprising that it resulted in poorer speech
perception. However, there appeared to be very little perceptual difference
between the two maps when subjects were wearing the CI together with their
HA/s. For word, consonant, and sentence recognition there were no significant
differences in patients’ speech scores between the place-matched and
conventional maps for the CIHA/s condition. In addition, subjects did not
indicate a clear subjective preference for either map. At 28 weeks postoperatively, subjects were given both maps to listen to and asked to complete
the APHAB questionnaire. Preference ratings from the questionnaire were small
and no trends were seen across the subject group. It seems likely that patients
- 201 were unable to determine a difference between the two maps in their everyday
listening environment.
In order to evaluate this further, confusion matrices were constructed of
subjects’ responses in the consonant test with both maps. These are shown in
Table 9.11 for the CI alone condition and in Table 9.12 for the CIHA/s condition.
For the CI alone condition, the place-matched map resulted in more overall
confusions than the conventional map. The most obvious of these was that
subjects confused the phoneme /d/ for /g/ with the place-matched map. The
phoneme /d/ was correctly recognized more often with the conventional map.
One exception to this trend was the phoneme /b/ which was recognized more
consistently with the place-matched map. For the conventional map /b/ was
often confused for /v/, /d/, or /g/. Although the frequency assignment of the
place-matched map varied between subjects, low-frequencies ranging from 250
up to 875 Hz were not represented by the CI processing. Considering the placematched map provided a smaller frequency range than the conventional fullrange map, it is not surprising that it resulted in more consonant confusions. A
different pattern can be seen when the two maps are compared in the CIHA/s
condition. The maps produced very similar results, although the phoneme /b/
was recognized correctly more frequently with the place-matched map.
Certainly, the addition of acoustic hearing for subjects in the current study gave
the listener the ability to combine the electric signal together with their natural
hearing. The CI stimulation representing high frequencies could be utilized
effectively by these listeners, regardless of the signal’s frequency range. The
overlap in frequency range of the conventional map together with the HA did not
negatively affect speech perception for these patients. This result supports
previous studies in which listeners appeared to adapt to frequency shifts with
experience and training (Dorman et al., 2003; Fu et al., 2003; Rosen et al.,
1999).
Stimulus
- 202 – Chapter 9 Combining electric and acoustic stimulation
p
t
k
b
d
g
m
n
s
sh
z
f
ch
j
th
v
p
t
k
b
d
g
m
n
s
sh
z
f
ch
j
th
v
p
15
t
3
23
4
19
26
p
17
t
26
1
18
1
27
Place-matched map (12 weeks post-operative)
k
b
d
g m n
s sh z
f ch
9
3
1
6
26
7
1
21
13 17
5
25
2
6
16
3
1
1
5 19
1
29
1
1 23
5
6
1 21
27
1
29
4
2
19
4
2
8
41 26 22 46 25 25 36 24 23 55 48
Conventional map (24 weeks post-operative)
k
b
d
g m n
s sh z
f ch
12
1
1
2
6
1
22
7
4
12
7
23
29
6
24
6 24
30
4
26
3
27
1
28
30
1
2
3
20
34
18
30
42
30
30
37
27
27
48
39
j
th
v
1
1
5
1
2
3
26
26
2
1
4
3
19
34
j
th
v
1
2
5
1
1
0
27
27
4
3
11
Table 9.11 Confusion matrices for the consonant test in the CI alone condition. The upper
panel shows subject responses when wearing the place-matched map. The lower panel
shows subject responses when wearing the conventional map. Correct responses for
each phoneme are shown in bold type diagonally. The total number of times each
phoneme was selected as a response is shown in bold type at the bottom of each matrix.
3
26
35
Stimulus
- 203 -
p
t
k
b
d
g
m
n
s
sh
z
f
ch
j
th
v
p
t
k
b
d
g
m
n
s
sh
z
f
ch
j
th
v
p
16
t
4
28
2
18
32
p
16
t
3
26
1
1
17
30
Place-matched map (12 weeks post-operative)
k
b
d
g m n
s sh z
f ch
9
1
1
1
1
2
25
3
15 10
1 22
7
4
26
3
26
10 14
3
2
30
1 25
4
1
29
4
26
30
1
6
4
21
1
5
1
39 17 33 38 36 14 40 25 40 50 43
Conventional map (24 weeks post-operative)
k
b
d
g m n
s sh z
f ch
11
4
28
3
8 11
1 23
6
30
4
26
6 24
29
1 25
4
1
29
25
2
28
1
3
29
3
39 9 34 40 32 28 31 27 29 57 39
j
th
v
2
1
1
23
23
4
2
6
1
21
26
j
th
v
1
7
1
5
25
25
Table 9.12 Confusion matrices for the consonant test in the CIHA/s condition. The upper
panel shows subject responses when wearing the place-matched map. The lower panel
shows subject responses when wearing the conventional map. Correct responses for
each phoneme are shown in bold type diagonally. The total number of times each
phoneme was selected as a response is shown in bold type at the bottom of each matrix.
1
1
3
7
24
36
- 204 – Chapter 9 Combining electric and acoustic stimulation
9.4.1 Conclusions
In summary, the effects of cochlear implantation were
investigated in 5 listeners with steeply-sloping losses. Their
hearing thresholds and speech perception in quiet and noise
were monitored for 28 weeks post-operatively. During this time,
subjects tried, in sequence, two maps with different frequency
ranges to determine whether the use of a frequency placematched map would provide any advantage over a conventional
full-range map. The results can be summarized as follows.
1. The average hearing threshold drop was 27 dB for the 4
subjects who retained some hearing thresholds post-operatively
in the implanted ear. One subject lost all hearing as a result of
the surgery.
2. For the group, cochlear implantation provided superior
recognition of monosyllabic words, consonants, and sentences
in noise compared to pre-operative scores.
3. In general, the best result in both quiet and noise was
obtained when subjects were wearing the implant together with
hearing aid/s.
4. No significant differences were found between the placematched map and the conventional full-range map for scores in
quiet or in noise.
- 205 -
Chapter 10
General conclusions
The current study examined aspects of severe high-frequency hearing loss.
Adult hearing-impaired listeners participated in three experiments to determine
the best way of providing high-frequency information: conventional
amplification, frequency compression, or cochlear implantation.
Conventional hearing devices aim to correct for threshold elevation in hearing
loss by presenting the listener with an amplified signal. Chapter 2 summarized
past studies which found that high-frequency amplification was not always
desirable. The condition with the most audible high-frequency signal resulted in
a worsening in speech understanding for some listeners. It has been proposed
that one possible reason for this lack of benefit was the presence of extensive
high-frequency dead regions (lack of functioning IHCs) in the cochlea.
The first aim of the current study was to evaluate whether high-frequency
amplification could provide benefit for listeners with moderate-to-profound
sloping hearing losses. Experiment 1 (Chapter 7) investigated this by testing 10
adult subjects with moderate-to-profound high-frequency hearing losses. A
clinical method for diagnosing dead regions known as the TEN test was carried
out with all subjects as well as a consonant test which included various lowpass filter conditions. Nine out of ten subjects showed an improvement in
speech scores with each increasing low-pass filter condition. Extensive dead
regions were found for one subject with the TEN test. For this subject, speech
perception results were consistent with the TEN test findings. The remaining
subjects may have had dead regions above 3000 Hz, because of the severity of
their hearing losses, but these could not be demonstrated with the TEN test.
Average consonant scores for the subject group improved significantly (p <
0.05) with increasing audibility of high-frequency components of the speech
signal. There were no cases of speech perception being reduced with
increasing bandwidth. In general, the results suggest that listeners with severe
- 206 – Chapter 10 General conclusions
high-frequency losses are often able to make some use of high-frequency
speech cues if these cues can be made audible. However, the levels of
audibility achieved in the experiments described above would be impractical for
everyday use with a conventional hearing aid. Acoustic feedback would limit the
amount of usable gain, dependent on individual characteristics, such as how
effectively the ear canal was sealed by the earmold. High levels of audibility
could also result in discomfort for the listener.
Therefore, alternative methods of presenting high-frequency information were
investigated. The second aim of the current study was to evaluate whether
useful high-frequency information could be provided to listeners with severe
high-frequency hearing loss by either a frequency-compression hearing aid or a
cochlear implant. Experiment 2 (Chapter 8) described how speech perception
testing in quiet and noise was carried out with a group of 7 adults with steeplysloping losses. Each of the subjects was fitted first with a conventional device,
followed by a frequency-compression device. They were required to wear each
device for 3-4 weeks in their everyday life. No significant differences in group
mean scores were found between the frequency-compression device and the
conventional device for understanding speech in quiet. Testing in noise showed
improvements for the frequency-compression scheme for only one of the five
subjects tested. Subjectively, all but one of the subjects showed higher scores
with the APHAB questionnaire for the conventional device.
Unfortunately, frequency compression did not prove to be effective at providing
high-frequency speech cues for this group of listeners. It was hoped that
cochlear implantation may provide more favourable results. Subjects who
participated in the frequency-compression trial were given the option to undergo
cochlear implantation (Experiment 3, Chapter 9). Three subjects who
participated in the frequency-compression trial opted to go ahead with cochlear
implantation. An additional two subjects were recruited. All five patients
obtained significant improvements in speech perception measures. On average,
phoneme scores at 12 weeks post-operatively improved from a pre-operative
score of 44% to a post-operative score of 66% with CI alone and 77% with
CIHA/s. Sentence understanding in noise at 12 weeks post-operatively showed
improvements from pre-operative scores for all of the four patients tested in the
- 207 CIHA/s condition. A case study of interest was S36. This patient was a non
hearing-aid user. Both the conventional and frequency-compression devices
were worn infrequently by the subject in everyday life. In contrast, S36 is highly
dependent on his cochlear implant and wears it consistently every day.
Subject numbers for all experiments were small which limits generalizations.
However, cochlear implantation appeared to be more effective than frequencycompression at providing improved speech perception for listeners with steeplysloping losses. Cochlear implantation does have some limitations. A loss of
hearing in the implanted ear, general risks of surgery, as well as the possibility
that the electrical stimulation would not provide a meaningful percept are some
of the risks the patient must consider. In addition, many patients may not be
willing or able to undergo surgery.
Hearing-aid technology is constantly being updated and improved. Patients
should be given access to all possible rehabilitative options. Processing such as
frequency-compression could certainly be considered. The clinician should be
aware of factors such as the frequency at which the patient’s useful hearing
ends as well as the steepness of the slope of the audiogram. The lower this
frequency, and the steeper the slope, the more unlikely it is that frequency
compression would be beneficial.
The third objective of the current study was to evaluate the effect of providing
electric hearing in regions of profound high-frequency sensorineural hearing
loss together with acoustic hearing in regions of normal-to-moderately impaired
low-frequency hearing in individuals with suitable types of hearing impairment.
Of the 5 patients who were implanted, the drop in hearing thresholds in the
implanted ear was on average 27 dB for four of the five patients. One patient
lost all hearing in the implanted ear. When compared to pre-operative scores,
for the word test, the group’s average phoneme score increased by 22
percentage points with CI alone and by 33 percentage points with CIHA/s at 12
weeks post-operatively. For the consonant test, the group’s mean scores
improved by 40 percentage points with CI alone, and 47 percentage points with
CIHA/s at 12 weeks post-operatively. For the two subjects who were fitted with
an ITE, no difference was found in scores between CIHAi, CIHAc, and CIHA/s.
In general, the best result in both quiet and noise was obtained when the
- 208 – Chapter 10 General conclusions
measurement was carried out with the subject wearing their HA/s together with
the CI. Low-frequency acoustic information appeared to benefit most patients.
However, it remains unclear what advantage CI recipients with low-frequency
acoustic hearing in both ears may have over CI recipients with acoustic hearing
in one ear only.
The final objective of the current study was to optimize the fitting of a cochlear
implant together with aided residual hearing, in particular by means of matching
frequency and/or perceived pitch between acoustic and electric modalities. A
place-matched map was created which the subject wore for the first 3 months
post-operatively. At this point the subject was switched to a conventional fullfrequency range map. No significant differences were found between the two
maps for testing carried out in quiet and in noise. In short, the input-frequency
range of the electric signal did not appear to have any significant impact on
patients’ scores when subjects were wearing both the CI together with HA/s in
these tests.
The following section will outline possible areas for future research.
Future research
The frequency-compression scheme tested in the current study showed
benefits for some adults with moderately-sloping losses. Reported benefits with
cochlear implantation in adult listeners with steeply-sloping losses have been
shown in numerous studies. Both frequency-compression and cochlear
implantation could be tested with additional hearing-impaired populations.
These could include children with similar hearing losses, or adults with varying
audiogram configurations.
Processing parameters continue to require refinement for both frequencycompression devices and cochlear implants. The cut-off frequency and slope of
the frequency compression could be more formally investigated when fitting the
device. Optimum values should be defined for the listener dependent on
individual characteristics, such as the audiogram. Once these parameters have
been set, it would be interesting to determine what effects the frequencycompression processing has on sound quality, music and sound localization
abilities. When fitting a cochlear implant together with a hearing aid in the same
- 209 ear, little is known about the effects of residual acoustic hearing in the implanted
ear. The combination of acoustic and electric hearing appears to provide
benefits over electric hearing alone. Further testing would need to be carried out
to determine how this hearing impacts on listening in noisy environments,
localization skills, or general sound quality. Further work is required to
determine the nature of the temporal and spectral information the acoustic
hearing gives the listener and how this can be best utilized together with electric
processing.
A large amount of existing research describing listeners with steeply-sloping
losses and cochlear implantation has been focused on how to preserve their
residual hearing. Electrode design and surgical modifications have been
explored. However, very little is known about how best to combine the
processing capabilities of both systems i.e. the speech processor and hearing
aid. The current study described the effects of varying the acoustic inputfrequency range of the CI. This is only one of many processing variables which
can be changed in the CI. The speech processing strategy, rate, and mode of
stimulation, electrode filter bandwidth, amplitude growth function and input
dynamic range may all have had some effect on patient’s scores. Similarly, a
different outcome may have occurred by changing the gain, compression ratio,
compression threshold, or number of channels in the hearing aid.
In conclusion, the combination of cochlear implants together with hearing aids is
proving to be an exciting new field of research. So far, it has been shown to be
an effective alternative rehabilitation for hearing-impaired listeners with steeplysloping losses. Future work should focus on optimizing design and fitting by
joint collaboration between both hearing aid and cochlear implant industries.
- 210 – Bibliography
Bibliography
ANSI. 1969. ANSI S3.5 - 1969. The calculation of the speech intelligibility index.
American National Standards Institute, New York.
ANSI. 1993. ANSI 53.9 draft v 3.1. American National Standards Institute, New York.
ANSI. 1997. ANSI S3.5 - 1997. The calculation of the speech intelligibility index.
American National Standards Institute, New York.
Armstrong, M., Pegg, P., James, C., Blamey, P. 1997. Speech perception in noise with
implant and hearing aid. American Journal of Otology 18, S140-1.
Baer, T., Moore, B.C., Kluk, K. 2002. Effects of low pass filtering on the intelligibility
of speech in noise for people with and without dead regions at high frequencies.
J Acoust Soc Am. 112, 1133-44.
Balkany, T., Luxford, W., Martinez, S., Hough, J.V.D. 1988. Surgical anatomy and
technique of cochlear implantation. In: Johnson, J.T., Blitzer, A., Ossoff, R.,
Thomas, J.R., (Eds.), Instructional courses. American Academy of
Otolaryngology - Head and neck surgery, Vol. 1. C.V. Mosby, St. Louis. pp.
195-218.
Baskent, D., Shannon, R.V. 2003. Speech recognition under conditions of frequencyplace compression and expansion. J Acoust Soc Am. 113, 2064-76.
Baskent, D., Shannon, R.V. 2004. Frequency-place compression and expansion in
cochlear implant listeners. J Acoust Soc Am. 116, 3130-40.
Baskent, D., Shannon, R.V. 2005. Interactions between cochlear implant electrode
insertion depth and frequency-place mapping. J Acoust Soc Am. 117, 1405-16.
Bennett, D.S., Byers, V.W. 1967. Increased intelligibility in the hypacusic by slow-play
frequency transposition. J Aud Res. 7, 107-118.
Bentler, R.A., Pavlovic, C.V. 1989. Transfer functions and correction factors used in
hearing aid evaluation and research. Ear Hear. 10, 58-63.
Bierer, J.A., Middlebrooks, J.C. 2002. Auditory cortical images of cochlear-implant
stimuli: dependence on electrode configuration. J Neurophysiol 87, 478-92.
Blamey, P.J., Dooley, G.J., Parisi, E.S., Clark, G.M. 1996. Pitch comparisons of
acoustically and electrically evoked auditory sensations. Hearing Research 99,
139-50.
Boggess, W.J., Baker, J.E., Balkany, T.J. 1989. Loss of residual hearing after cochlear
implantation. Laryngoscope 99, 1002-5.
Boothroyd, A., Hanin, L., Hnath, T. 1985. A sentence test of speech perception:
reliability, set equivalence, and short term learning. Internal report RCI 10,
Speech and Hearing Sciences Research Centre, City University of New York.
Borg, E., Canlon, B., & Engstrom, B. 1995. Noise-induced hearing loss - literature
review and experiments in rabbits. Scand Audiol. Supp. 40, 1-147.
Bredberg, G. 1968. Cellular pattern and nerve supply of the human organ of Corti. Acta
Otolaryngol, Suppl 236:1+.
- 211 Byrne, D., Dillon, H. 1986. The National Acoustic Laboratories' (NAL) new procedure
for selecting the gain and frequency response of a hearing aid. Ear Hear. 7, 25765.
Byrne, D., Dillon, H., Ching, T., Katsch, R., Keidser, G. 2001. NAL-NL1 procedure for
fitting nonlinear hearing aids: characteristics and comparisons with other
procedures. J Am Acad Audiol. 12, 37-51.
Byrne, D., Dillon, H., Tran, K., Arlinger, S., Wilbraham, K., Cox, R.M., Hagerman, B.,
Hetu, R., Kei, J., Lui, C., Kiessling, J., Nasser Kotby, M., Nasser, H.A., El
Kholy, W., Nakanishi, Y., Oyer, H., Powell, R., Stephens, D., Meredith, R.,
Sirimanna, T., Tavartkiladze, G., Frolenkov, G., Westerman, S., Ludvigsen, C.
1994. An international comparison of long-term average speech spectra. J
Acoust Soc Am. 96, 2108-2120.
Carhart, R., Jerger, J.F. 1959. Preferred method for clinical determination of pure-tone
thresholds. J Speech Hear Disord. 24, 330-345.
Ching, T.Y., Dillon, H., Byrne, D. 1998. Speech recognition of hearing-impaired
listeners: predictions from audibility and the limited role of high-frequency
amplification. J Acoust Soc Am. 103, 1128-40.
Ching, T.Y., Incerti, P., Hill, M. 2004. Binaural benefits for adults who use hearing aids
and cochlear implants in opposite ears. Ear Hear. 25, 9-21.
Ching, T.Y., Psarros, C., Hill, M., Dillon, H., Incerti, P. 2001. Should children who use
cochlear implants wear hearing aids in the opposite ear? Ear Hear. 22, 365-80.
Cohen, L.T., Busby, P.A., Whitford, L.A., Clark, G.M. 1996. Cochlear implant place
psychophysics 1. Pitch estimation with deeply inserted electrodes. Audiol
Neurootol 1, 265-77.
Cox, R.M., Alexander, G.C. 1995. The abbreviated profile of hearing aid benefit. Ear
Hear. 16, 176-86.
David, E.E., Ostroff, J.M., Shipp, D., Nedzelski, J.M., Chen, J.M., Parnes, L.S.,
Zimmerman, K., Schramm, D., Seguin, C. 2003. Speech coding strategies and
revised cochlear implant candidacy: an analysis of post-implant performance.
Otol Neurotol 24, 228-33.
Davis-Penn, W., Ross, M. 1993. Pediatric experiences with frequency transposing.
Hearing Instruments 44, 26-32.
Deman, P.R., van Dijk, B., Offeciers, F.E., Govaerts, P.J. 2004. Pitch estimation of a
deeply inserted cochlear implant electrode. Int J Audiol. 43, 363-8.
Dettman, S.J., D'Costa, W.A., Dowell, R.C., Winton, E.J., Hill, K.L., Williams, S.S.
2004. Cochlear implants for children with significant residual hearing. Arch
Otolaryngol Head Neck Surg 130, 612-8.
Dooley, G.J., Blamey, P.J., Seligman, P.M., Alcantara, J.I., Clark, G.M., Shallop, J.K.,
Arndt, P., Heller, J.W., Menapace, C.M. 1993. Combined electrical and
acoustical stimulation using a bimodal prosthesis. Arch Otolaryngol Head Neck
Surg 119, 55-60.
Dorman, M.F., Ketten, D. 2003. Adaptation by a cochlear-implant patient to upward
shifts in the frequency representation of speech. Ear Hear 24, 457-60.
- 212 – Bibliography
Dorman, M.F., Loizou, P.C., Rainey, D. 1997. Simulating the effect of cochlear-implant
electrode insertion depth on speech understanding. J Acoust Soc Am. 102, 29936.
Dowell, R.C., Hollow, R., Winton, E. 2004. Outcomes for cochlear implant users with
significant residual hearing: implications for selection criteria in children. Arch
Otolaryngol Head Neck Surg 130, 575-81.
Dreschler, W.A., Verschuure, H., Ludvigsen, C., Westermann, S. 2001. ICRA noises:
artificial noise signals with speech-like spectral and temporal properties for
hearing instrument assessment. International Collegium for Rehabilitative
Audiology. Audiology 40, 148-57.
Dubno, J.R., Dirks, D.D., Langhofer, L.R. 1982. Evaluation of hearing-impaired
listeners using a nonsense-syllable test. II. Syllable recognition and consonant
confusion patterns. J Speech Hear Res. 25, 141-8.
Dubno, J.R., Dirks, D.D., Ellison, D.E. 1989a. Stop-consonant recognition for normalhearing listeners and listeners with high-frequency hearing loss. I: The
contribution of selected frequency regions. J Acoust Soc Am. 85, 347-54.
Dubno, J.R., Dirks, D.D., Schaefer, A.B. 1989b. Stop-consonant recognition for
normal-hearing listeners and listeners with high-frequency hearing loss. II:
Articulation index predictions. J Acoust Soc Am. 85, 355-64.
Engstrom, B. 1983. Stereocilia of sensory cells in normal and hearing impaired ears.
Scand Audiol. Suppl. 19, 1-34.
Fletcher, H. 1953. Speech and hearing in communication. Van Nostrand, New York.
Fletcher, H., Galt, R.H. 1950. The perception of speech and its relation to telephony. J
Acoust Soc Am. 22, 89-151.
French, N.R., Steinberg, J.C. 1947. Factors governing the Intelligibility of speech
sounds. J Acoust Soc Am. 19, 90-119.
Friesen, L.M., Shannon, R.V., Baskent, D., Wang, X. 2001. Speech recognition in noise
as a function of the number of spectral channels: comparison of acoustic hearing
and cochlear implants. J Acoust Soc Am. 110, 1150-63.
Fu, Q.J., Shannon, R.V. 1999. Recognition of spectrally degraded and frequency-shifted
vowels in acoustic and electric hearing. J Acoust Soc Am. 105, 1889-900.
Fu, Q.J., Galvin, J.J., 3rd. 2003. The effects of short-term training for spectrally
mismatched noise-band speech. J Acoust Soc Am. 113, 1065-72.
Fu, Q.J., Shannon, R.V., Wang, X. 1998. Effects of noise and spectral resolution on
vowel and consonant recognition: acoustic and electric hearing. J Acoust Soc
Am. 104, 3586-96.
Fu, Q.J., Shannon, R.V., Galvin, J.J., 3rd. 2002. Perceptual learning following changes
in the frequency-to-electrode assignment with the Nucleus-22 cochlear implant.
J Acoust Soc Am. 112, 1664-74.
Gantz, B.J., Turner, C.W. 2003. Combining acoustic and electrical hearing.
Laryngoscope 113, 1726-30.
Gantz, B.J., Turner, C. 2004. Combining acoustic and electrical speech processing:
Iowa/Nucleus hybrid implant. Acta Oto-Laryngol. 124, 344-7.
- 213 Gantz, B.J., Turner, C., Gfeller, K.E., Lowder, M.W. 2005. Preservation of hearing in
cochlear implant surgery: advantages of combined electrical and acoustical
speech processing. Laryngoscope 115, 796-802.
Gantz, B.J., Tyler, R.S., Knutson, J.F., Woodworth, G., Abbas, P., McCabe, B.F.,
Hinrichs, J., Tye-Murray, N., Lansing, C., Kuk, F., et al. 1988. Evaluation of
five different cochlear implant designs: audiologic assessment and predictors of
performance. Laryngoscope 98, 1100-6.
Gelfand, S.A. 1998. Hearing: an introduction to psychological and physiological
acoustics Marcel Dekker, Inc., New York, NY.
Greenwood, D.D. 1990. A cochlear frequency-position function for several species--29
years later. J Acoust Soc Am. 87, 2592-605.
Gstoettner, W., Franz, P., Hamzavi, J., Plenk, H., Jr., Baumgartner, W., Czerny, C.
1999. Intracochlear position of cochlear implant electrodes. Acta Otolaryngol
119, 229-33.
Gstoettner, W., Kiefer, J., Baumgartner, W.D., Pok, S., Peters, S., Adunka, O. 2004.
Hearing preservation in cochlear implantation for electric acoustic stimulation.
Acta Oto-Laryngol. 124, 348-52.
Gstoettner, W.K., Hamzavi, J., Baumgartner, W.D. 1998. Speech discrimination scores
of postlingually deaf adults implanted with the Combi 40 cochlear implant. Acta
Otolaryngol 118, 640-5.
Hamzavi, J., Pok, S.M., Gstoettner, W., Baumgartner, W.D. 2004. Speech perception
with a cochlear implant used in conjunction with a hearing aid in the opposite
ear. International Journal of Audiology 43, 61-5.
Hamzavi, J., Baumgartner, W.D., Pok, S.M., Franz, P., Gstoettner, W. 2003. Variables
affecting speech perception in postlingually deaf adults following cochlear
implantation. Acta Otolaryngol 123, 493-8.
Hawkins, D.B., Walden, B.E., Montgomery, A., Prosek, R.A. 1987. Description and
validation of an LDL procedure designed to select SSPL90. Ear Hear. 8, 162-9.
Helms, J., Muller, J., Schon, F., Moser, L., Arnold, W., Janssen, T., Ramsden, R., von
Ilberg, C., Kiefer, J., Pfennigdorf, T., Gstottner, W., Baumgartner, W.,
Ehrenberger, K., Skarzynski, H., Ribari, O., Thumfart, W., Stephan, K., Mann,
W., Heinemann, M., Zorowka, P., Lippert, K.L., Zenner, H.P., Bohndord, M.,
Huttenbrink, K., Hochmair-Desoyer, I., et al. 1997. Evaluation of performance
with the COMBI40 cochlear implant in adults: a multicentric clinical study.
ORL J Otorhinolaryngol Relat Spec 59, 23-35.
Hochberg, Y., Tamhane, A.C. 1987. Multiple Comparison Procedures Wiley, Hobeken,
NJ.
Hochmair-Desoyer, I.J., Hochmair, E.S., Burian, K., Stiglbrunner, H.K. 1983. Percepts
from the Vienna cochlear prosthesis. Ann N Y Acad Sci 405, 295-306.
Hodges, A.V., Schloffman, J., Balkany, T. 1997. Conservation of residual hearing with
cochlear implantation. American Journal of Otology 18, 179-83.
Hogan, C.A., Turner, C.W. 1998. High-frequency audibility: benefits for hearingimpaired listeners. Journal of the Acoustical Society of America 104, 432-41.
- 214 – Bibliography
Hollow, R., Salomon, R., Tselepis, V., Leigh, J., Dowell, R.C. 2006. An up-date on
speech recognition benefits for adults using cochlear implants. Australian and
New Zealand Journal of Audiology 28, 27.
Hornsby, B.W., Ricketts, T.A. 2003. The effects of hearing loss on the contribution of
high- and low-frequency speech information to speech understanding. J Acoust
Soc Am. 113, 1706-17.
House, W.F., Berliner, K.I., Eisenberg, L.S., Edgerton, B.J., Thielemeir, M.A. 1981.
The cochlear implant: 1980 update. Acta Otolaryngol 91, 457-62.
Humes, L.E., Riker, S. 1992. Evaluation of two clinical versions of the articulation
index. Ear Hear. 13, 406-9.
Humes, L.E., Dirks, D.D., Bell, T.S., Ahlstrom, C., Kincaid, G.E. 1986. Application of
the Articulation Index and the Speech Transmission Index to the recognition of
speech by normal-hearing and hearing-impaired listeners. J Speech Hear Res.
29, 447-62.
James, C., Albegger, K., Battmer, R., Burdo, S., Deggouj, N., Deguine, O., Dillier, N.,
Gersdorff, M., Laszig, R., Lenarz, T., Rodriguez, M.M., Mondain, M., Offeciers,
E., Macias, A.R., Ramsden, R., Sterkers, O., Von Wallenberg, E., Weber, B.,
Fraysse, B. 2005. Preservation of residual hearing with cochlear implantation:
how and why. Acta Otolaryngol 125, 481-91.
Johansson, B. 1961. A new coding amplifier system for the severely hard of hearing.
Proceedings of the 3rd International Congress on Acoustics, Stuttgart 1959 2,
655-657.
Kamm, C.A., Dirks, D.D., Bell, T.S. 1985. Speech recognition and the Articulation
Index for normal and hearing-impaired listeners. J Acoust Soc Am. 77, 281-8.
Kawano, A., Seldon, H.L., Clark, G.M. 1996. Computer-aided three-dimensional
reconstruction in human cochlear maps: measurement of the lengths of organ of
Corti, outer wall, inner wall, and Rosenthal's canal. The Annals of otology,
rhinology, and laryngology 105, 701-9.
Kessler, D.K. 1999. The CLARION multi-strategy cochlear implant. Ann Otol Rhinol
Laryngol 108 Suppl 177, 8-16.
Ketten, D.R., Skinner, M.W., Wang, G., Vannier, M.W., Gates, G.A., Neely, J.G. 1998.
In vivo measures of cochlear length and insertion depth of nucleus cochlear
implant electrode arrays. Ann Otol Rhinol Laryngol Suppl 175, 1-16.
Khanna, S.M., Leonard, D.G. 1982. Basilar membrane tuning in the cat cochlea.
Science 215, 305-6.
Khanna, S.M., Leonard, D.G. 1986a. Relationship between basilar membrane tuning
and hair cell condition. Hearing Research 23, 55-70.
Khanna, S.M., Leonard, D.G. 1986b. Measurement of basilar membrane vibrations and
evaluation of the cochlear condition. Hearing Research 23, 37-53.
Kiang, N.Y. 1965. Discharge patterns of single fibers in the cat's auditory nerve M.I.T.
Press, Cambridge, Mass.
Kiefer, J., Hohl, S., Sturzebecher, E., Pfennigdorff, T., Gstoettner, W. 2001.
Comparison of speech recognition with different speech coding strategies
(SPEAK, CIS, and ACE) and their relationship to telemetric measures of
- 215 compound action potentials in the nucleus CI 24M cochlear implant system.
Audiology 40, 32-42.
Kiefer, J., von Ilberg, C., Reimer, B., Knecht, R., Gall, V., Diller, G., Sturzebecher, E.,
Pfennigdorff, T., Spelsberg, A. 1998. Results of cochlear implantation in
patients with severe to profound hearing loss--implications for patient selection.
Audiology 37, 382-95.
Kiefer, J., Pok, M., Adunka, O., Sturzebecher, E., Baumgartner, W., Schmidt, M.,
Tillein, J., Ye, Q., Gstoettner, W. 2005. Combined electric and acoustic
stimulation of the auditory system: results of a clinical study. Audiol Neorootol.
10, 134-44.
Kileny, P.R., Zwolan, T.A., Telian, S.A., Boerst, A. 1998. Performance with the 20 +
2L lateral wall cochlear implant. Am J Otol 19, 313-9.
Koch, D.B., Osberger, M.J., Segel, P., Kessler, D. 2004. HiResolution and conventional
sound processing in the HiResolution bionic ear: using appropriate outcome
measures to assess speech recognition ability. Audiol Neurootol. 9, 214-23.
Leake, P.A., Stakhovskaya, O., Sridhar, D. 2005. Frequency map for the human spiral
ganglion, Conference on Implantable Auditory Prostheses, Asilomar, Pacific
Grove, CA.
Lehnhart, E. 1993. Intracochlear placement of the cochlear implant electrodes in soft
surgery technique. HNO, 356-359.
Liberman, M.C. 1984. Single-neuron labeling and chronic cochlear pathology. I.
Threshold shift and characteristic-frequency shift. Hearing Research 16, 33-41.
Liberman, M.C., Kiang, N.Y. 1984a. Single-neuron labeling and chronic cochlear
pathology. IV. Stereocilia damage and alterations in rate- and phase-level
functions. Hearing Research 16, 75-90.
Liberman, M.C., Dodds, L.W. 1984b. Single-neuron labeling and chronic cochlear
pathology. III. Stereocilia damage and alterations of threshold tuning curves.
Hearing Research 16, 55-74.
Liberman, M.C., Dodds, L.W. 1984c. Single-neuron labeling and chronic cochlear
pathology. II. Stereocilia damage and alterations of spontaneous discharge rates.
Hearing Research 16, 43-53.
Liberman, M.C., Oliver, M.E. 1984d. Morphometry of intracellularly labeled neurons of
the auditory nerve: correlations with functional properties. The Journal of
Comparative Neurology 223, 163-76.
Liberman, M.C., Dodds, L.W., Learson, D.A. 1986. Structure-function correlation in
noise-damaged ears: a light and electron-microscopic study Plenum Press, New
York.
Lie, K.F. 2003. Evaluating the presence of residual hearing in adult cochlear implant
users: Minor Thesis, Department of Otolaryngolgy, The University of
Melbourne.
Lim, D.J. 1980. Cochlear anatomy related to cochlear micromechanics. A review. J
Acoust Soc Am. 67, 1686-95.
Lim, D.J. 1986. Functional structure of the organ of Corti: a review. Hearing Research
22, 117-46.
- 216 – Bibliography
Ling, D. 1968. Three experiments on frequency transposition. American Annals of the
Deaf 113, 283-294.
Mackersie, C.L., Crocker, T.L., Davis, R.A. 2004. Limiting high-frequency hearing aid
gain in listeners with and without suspected cochlear dead regions. J Am Acad
Audiol. 15, 498-507.
Markham, D., Hazan, V. 2002. The UCL speaker database. Speech, Hearing and
Language: UCL Work in Progress 14, 1-17.
Marsh, M.A., Xu, J., Blamey, P.J., Whitford, L.A., Xu, S.A., Silverman, J.M., Clark,
G.M. 1993. Radiologic evaluation of multichannel intracochlear implant
insertion depth. Am J Otol 14, 386-91.
McDermott, H.J. 2004. Music perception with cochlear implants: a review. Trends
Amplif. 8, 49-82.
McDermott, H.J., Dean, M.R. 2000. Speech perception with steeply sloping hearing
loss: effects of frequency transposition. Br J Audiol. 34, 353-61.
McDermott, H.J., Knight, M.R. 2001. Preliminary results with the AVR ImpaCt
frequency-transposing hearing aid. J Am Acad Audiol. 12, 121-7.
McDermott, H.J., Dorkos, V.P., Dean, M.R., Ching, T.Y. 1999. Improvements in
speech perception with use of the AVR TranSonic frequency-transposing
hearing aid. J Speech Lang Hear Res. 42, 1323-35.
McDermott, H.J., Vandali, A.E., van Hoesel, R.J.M., McKay, C.M., Harrison, J.M.,
Cohen, L.T. 1993. A portable programmable digital sound processor for
cochlear implant research. IEEE Transactions on Rehabilitation Engineering 1,
94-100.
Miller, G., Nicely, P. 1955. An analysis of perceptual confusions among some English
consonants. J Acoust Soc Am. 27, 623- 637.
Mok, M., Grayden, D., Dowell, R.C., Lawrence, D. 2006. Speech perception for adults
who use hearing AIDS in conjunction with cochlear implants in opposite ears. J
Speech Lang Hear Res 49, 338-51.
Moore, B.C. 1978. Psychophysical tuning curves measured in simultaneous and forward
masking. J Acoust Soc Am. 63, 524-32.
Moore, B.C. 2001. Dead regions in the cochlea: diagnosis, perceptual consequences,
and implications for the fitting of hearing aids. Trends Amplif. 5, 1-34.
Moore, B.C. 2002. Response to "Articulation index predictions for hearing-impaired
listeners with and without cochlear dead regions". J Acoust Soc Am. 111, 25492550.
Moore, B.C. 2004. Dead regions in the cochlea: conceptual foundations, diagnosis, and
clinical applications. Ear Hear 25, 98-116.
Moore, B.C., Glasberg, B.R. 1998. Use of a loudness model for hearing-aid fitting. I.
Linear hearing aids. Br J Audiol. 32, 317-35.
Moore, B.C., Alcantara, J.I. 2001. The use of psychophysical tuning curves to explore
dead regions in the cochlea. Ear Hear. 22, 268-78.
Moore, B.C., Huss, M., Vickers, D.A., Glasberg, B.R., Alcantara, J.I. 2000a. A test for
the diagnosis of dead regions in the cochlea. Br J Audiol. 34, 205-24.
- 217 Moore, B.C.J., Glasberg, B.R. 1997. A model of loudness perception applied to
cochlear hearing loss. Aud Neurosci. 3, 289-311.
Moore, B.C.J., Huss, M., Glasberg, B.R. 2000b. Diagnosis of dead regions. Starkey
Laboratories, Inc., Eden Prairie, MN.
Moore, B.C.J., Glasberg, B.R., Stone, M.A. 2003. The TEN (HL) test for diagnosis of
dead regions in the cochlea. Department of Experimental Psychology,
University of Cambridge, UK.
Muller, J., Schon, F., Helms, J. 2002. Speech understanding in quiet and noise in
bilateral users of the MED-EL COMBI 40/40+ cochlear implant system. Ear
Hear 23, 198-206.
Murray, N., Byrne, D. 1986. Performance of hearing-impaired and normal hearing
listeners with various high frequency cut-offs in hearing aids. Aust J Audiol. 8,
21-28.
National Acoustic Laboratories. 2000. Speech and noise for hearing aid evaluation.
Australian Hearing, Chatswood, NSW.
Neary, T. 1989. Static, dynamic, and relational properties in vowel perception. J Acoust
Soc Am. 85, 2088-2113.
Nelson, P.B., Jin, S.H., Carney, A.E., Nelson, D.A. 2003. Understanding speech in
modulated interference: cochlear implant users and normal-hearing listeners. J
Acoust Soc Am. 113, 961-8.
Parent, T.C., Chmiel, R., Jerger, J. 1998. Comparison of performance with frequency
transposition hearing aids and conventional hearing aids. J Am Acad Audiol. 9,
67-77.
Pavlovic, C.V. 1984. Use of the articulation index for assessing residual auditory
function in listeners with sensorineural hearing impairment. J Acoust Soc Am.
75, 1253-8.
Pavlovic, C.V., Studebaker, G.A. 1984. An evaluation of some assumptions underlying
the articulation index. J Acoust Soc Am. 75, 1606-12.
Peterson, G.E., Lehiste, I. 1962. Revised CNC lists for auditory tests. J Speech Hear
Disord. 27, 62-70.
Pfingst, B.E., Zwolan, T.A., Holloway, L.A. 1997. Effects of stimulus configuration on
psychophysical operating levels and on speech recognition with cochlear
implants. Hear Res 112, 247-60.
Pfingst, B.E., Franck, K.H., Xu, L., Bauer, E.M., Zwolan, T.A. 2001. Effects of
electrode configuration and place of stimulation on speech perception with
cochlear prostheses. J Assoc Res Otolaryngol 2, 87-103.
Posen, M.P., Reed, C.M., Braida, L.D. 1993. Intelligibility of frequency-lowered speech
produced by a channel vocoder. Journal of Rehabilitation and Research
Development 30, 26-38.
Rankovic, C.M. 1991. An application of the articulation index to hearing aid fitting. J
Speech Hear Res. 34, 391-402.
- 218 – Bibliography
Rankovic, C.M. 2002. Articulation index predictions for hearing-impaired listeners with
and without cochlear dead regions. J Acoust Soc Am. 111, 2545-8; author reply
2549-50.
Reed, C.M., Hicks, B.L., Braida, L.D., Durlach, N.I. 1983. Discrimination of speech
processed by low-pass filtering and pitch-invariant frequency lowering. J Acoust
Soc Am. 74, 409-19.
Reed, C.M., Schultz, K.I., Braida, L.D., Durlach, N.I. 1985. Discrimination and
identification of frequency-lowered speech in listeners with high-frequency
hearing impairment. J Acoust Soc Am. 78, 2139-41.
Rizer, F.M., Arkis, P.N., Lippy, W.H., Schuring, A.G. 1988. A postoperative
audiometric evaluation of cochlear implant patients. Otolaryngology Head Neck
Surgery 98, 203-6.
Rosen, S., Faulkner, A., Wilkinson, L. 1999. Adaptation by normal listeners to upward
spectral shifts of speech: implications for cochlear implants. J Acoust Soc Am
106, 3629-36.
Rosenhouse, J. 1990. A new transposition device for the deaf. The Hearing Journal 43,
20-25.
Rubinstein, J.T., Parkinson, W.S., Tyler, R.S., Gantz, B.J. 1999. Residual speech
recognition and cochlear implant performance: effects of implantation criteria.
Am J Otol 20, 445-52.
Ruggero, M.A. 1992. Responses to sound of the basilar membrane of the mammalian
cochlea. Curr Opin Neurobiol 2, 449-56.
Ruggero, M.A., Rich, N.C., Recio, A., Narayan, S.S., Robles, L. 1997. Basilarmembrane responses to tones at the base of the chinchilla cochlea. J Acoust Soc
Am. 101, 2151-63.
Seligman, P., McDermott, H. 1995. Architecture of the Spectra 22 speech processor.
Annals of Otology, Rhinology and Laryngology, Supplements 104, 139-41.
Sellick, P.M., Patuzzi, R., Johnstone, B.M. 1982. Modulation of responses of spiral
ganglion cells in the guinea pig cochlea by low frequency sound. Hearing
Research 7, 199-221.
Shallop, J.K., Arndt, P., Turnacliff, K.A. 1992. Expanded indications for cochlear
implantation: perceptual results in seven adults with residual hearing. Journal of
Speech-Language Pathology and Audiology 16, 141-148.
Shin, Y.J., Deguine, O., Laborde, M.L., Fraysse, B. 1997. [Conservation of residual
hearing after cochlear implantation]. Rev Laryngol Otol Rhinol (Bord) 118, 2338.
Simpson, A., McDermott, H.J., Dowell, R.C. 2005a. Benefits of audibility for listeners
with severe high-frequency hearing loss. Hearing Research 210, 42-52.
Simpson, A., Hersbach, A.A., McDermott, H.J. 2005b. Relationship between speech
recognition and self-report measures. Acoustics Australia 33, 57-62.
Simpson, A., Hersbach, A.A., McDermott, H.J. 2005c. Improvements in speech
perception with an experimental nonlinear frequency-compression hearing
device. Int J Audiol. 44, 281-292.
- 219 Simpson, A., Hersbach, A.A., McDermott, H.J. 2006. Frequency compression outcomes
for listeners with steeply sloping audiograms. Int J Audiol. 45, 619 - 629.
Skarzynski, H., Lorens, A., Piotrowska, A. 2003. A new method of partial deafness
treatment. Med Sci Monit 9, CS20-4.
Skarzynski, H., Lorens, A., D'Haese, P., Walkowiak, A., Piotrowska, A., Sliwa, L.,
Anderson, I. 2002. Preservation of residual hearing in children and postlingually deafened adults after cochlear implantation: an initial study. ORL;
Journal for Oto-rhino-laryngology and its Related Specialties 64, 247-53.
Skinner, M.W., Holden, L.K., Holden, T.A., Demorest, M.E., Fourakis, M.S. 1997.
Speech recognition at simulated soft, conversational, and raised-to-loud vocal
efforts by adults with cochlear implants. J Acoust Soc Am. 101, 3766-82.
Skinner, M.W., Holden, L.K., Whitford, L.A., Plant, K.L., Psarros, C., Holden, T.A.
2002. Speech recognition with the nucleus 24 SPEAK, ACE, and CIS speech
coding strategies in newly implanted adults. Ear Hear 23, 207-23.
Slepecky, N., Ulfendahl, M., Flock, A. 1988. Effects of caffeine and tetracaine on outer
hair cell shortening suggest intracellular calcium involvement. Hearing Research
32, 11-21.
Steinberg, J.C., Gardner, M.B. 1937. The dependence of hearing impairment on sound
intensity. J Acoust Soc Am. 9, 11-23.
Stelmachowicz, P.G., Pittman, A.L., Hoover, B.M., Lewis, D.E. 2001. Effect of
stimulus bandwidth on the perception of /s/ in normal- and hearing-impaired
children and adults. J Acoust Soc Am. 110, 2183-90.
Summers, V., Molis, M.R., Musch, H., Walden, B.E., Surr, R.K., Cord, M.T. 2003.
Identifying dead regions in the cochlea: psychophysical tuning curves and tone
detection in threshold-equalizing noise. Ear Hear 24, 133-42.
Thornton, A.R., Abbas, P.J. 1980. Low-frequency hearing loss: perception of filtered
speech, psychophysical tuning curves, and masking. J Acoust Soc Am. 67, 63843.
Turner, C., Burns, E.M., Nelson, D.A. 1983. Pure tone pitch perception and lowfrequency hearing loss. J Acoust Soc Am. 73, 966-75.
Turner, C.W., Robb, M.P. 1987. Audibility and recognition of stop consonants in
normal and hearing-impaired subjects. J Acoust Soc Am. 81, 1566-73.
Turner, C.W., Hurtig, R.R. 1999. Proportional frequency compression of speech for
listeners with sensorineural hearing loss. J Acoust Soc Am. 106, 877-86.
Turner, C.W., Henry, B.A. 2002. Benefits of amplification for speech recognition in
background noise. J Acoust Soc Am. 112, 1675-80.
Turner, C.W., Gantz, B.J., Vidal, C., Behrens, A., Henry, B.A. 2004. Speech
recognition in noise for cochlear implant listeners: benefits of residual acoustic
hearing. J Acoust Soc Am. 115, 1729-35.
Ulehlova, L., Voldrich, L., Janisch, R. 1987. Correlative study of sensory cell density
and cochlear length in humans. Hearing Research 28, 149-51.
van den Honert, C., Stypulkowski, P.H. 1987. Single fiber mapping of spatial excitation
patterns in the electrically stimulated auditory nerve. Hear Res 29, 195-206.
- 220 – Bibliography
Vandali, A.E., Whitford, L.A., Plant, K.L., Clark, G.M. 2000a. Speech perception as a
function of electrical stimulation rate: using the Nucleus 24 cochlear implant
system. Ear Hear 21, 608-24.
Vandali, A.E., Whitford, L.A., Plant, K.L., Clark, G.M. 2000b. Speech perception as a
function of electrical stimulation rate: using the Nucleus 24 cochlear implant
system. Ear Hear. 21, 608-24.
Velmans, M. 1973. Speech imitation in simulated deafness, using visual cues and
'recoded' auditory information. Lang Speech. 16, 224-36.
Velmans, M. 1974. The design of speech recoding devices for the deaf. Br J Audiol. 8,
1-5.
Velmans, M. 1975. Effects of frequency "recoding" on the articulation learning of
perceptively deaf children. Lang Speech. 18, 180-93.
Velmans, M., Marcuson, M. 1983. The acceptability of spectrum-preserving and
spectrum-destroying transposition to severely hearing-impaired listeners. Br J
Audiol. 17, 17-26.
Vestergaard, M.D. 2003. Dead regions in the cochlea: implications for speech
recognition and applicability of articulation index theory. Int J Audiol. 42, 24961.
Vickers, D.A., Baer, T., Moore, B.C.J. 2001. Effects of lowpass filtering on speech
intelligibility for listeners with dead regions at high frequenies. Br J Audiol. 35,
148-149.
von Ilberg, C., Kiefer, J., Tillein, J., Pfenningdorff, T., Hartmann, R., Sturzebecher, E.,
Klinke, R. 1999. Electric-acoustic stimulation of the auditory system. New
technology for severe hearing loss. ORL; Journal for Oto-rhino-laryngology and
its Related Specialties 61, 334-40.
Wedenberg, E. 1961. Auditory training of the severly hard of hearing using a coding
amplifier. Proceedings of the 3rd International Congress on Acoustics, Stuttgart
1959 2, 658-660.
Wilson, B.S. 2006. Speech processing strategies. In: Cooper, H.R., Craddock, L.C.,
(Eds.), Cochlear implants: A preactical guide, second ed. Whurr Publishers,
West Sussex, England.
Wilson, B.S., Finley, C.C., Lawson, D.T., Wolford, R.D., Eddington, D.K., Rabinowitz,
W.M. 1991. Better speech recognition with cochlear implants. Nature 352, 2368.
Wright, A. 1981. Scanning electron microscopy of the normal human cochlea. Clinical
Otolaryngology 6, 237-44.
Xu, J., Xu, S.A., Cohen, L.T., Clark, G.M. 2000. Cochlear view: postoperative
radiography for cochlear implantation. Am J Otol 21, 49-56.
Zakis, J.A., McDermott, H.J., Fisher, M. 2004. An advanced sound processing system
for hearing aid research. Iran J Audiol. 3, 75-78.
Zappia, J.J., Niparko, J.K., Oviatt, D.L., Kemink, J.L., Altschuler, R.A. 1991.
Evaluation of the temporal bones of a multichannel cochlear implant patient.
The Annals of Otology, Rhinology, and Laryngology 100, 914-21.
- 221 Ziese, M., Stutzel, A., von Specht, H., Begall, K., Freigang, B., Sroka, S., Nopp, P.
2000. Speech understanding with the CIS and the n-of-m strategy in the MEDEL COMBI 40+ system. ORL J Otorhinolaryngol Relat Spec 62, 321-9.
Zurek, P.M., Delhorne, L.A. 1987. Consonant reception in noise by listeners with mild
and moderate sensorineural hearing impairment. J Acoust Soc Am. 82, 1548-59.
- 222 – Bibliography
- 223 -
Appendices
Appendix A
Written information regarding experiments 1-3 given to each subject.
Appendix B Publications
1. Simpson, A., Hersbach, A.A., McDermott, H.J., 2005a. Improvements
in speech perception with an experimental nonlinear frequency-compression
hearing device. Int J Audiol., 44, 281-292.
2. Simpson, A., McDermott, H.J., Dowell, R.C., 2005c. Benefits of
audibility for listeners with severe high-frequency hearing loss. Hearing
Research, 210, 42-52.
3. Simpson, A., Hersbach, A.A., McDermott, H.J., 2006. Frequency
compression outcomes for listeners with steeply sloping audiograms. Int J
Audiol., 45, 619 - 629.
- 224 –Appendix A
- 225 THE ROYAL VICTORIAN EYE AND EAR HOSPITAL
EXPERIMENTAL SUBJECT’S STATEMENT OF RIGHTS
The Royal Victorian Eye and Ear Hospital considers it important that you know:
Any patient who is asked to participate in a research study involving medical experiment,
or who is requested to consent on behalf of another, has the right to:
1.
Be informed of the nature and purpose of the experiment.
2.
Be given an explanation of the procedures to be followed and any drugs used in
the medical experiment.
3.
Be given a description of discomforts and risks reasonably expected from the
experiment, if applicable.
4.
Be given an explanation of any benefits to the subject reasonably to be expected
from the experiment, if applicable.
5.
Be advised of appropriate, alternative procedures, drugs, or devices that might
be advantageous to the subject, and their relative risks and benefits.
6.
Be informed of the avenue of medical treatment, if any, available to the subject
after the experiment if complications should arise.
7.
Be given an opportunity to ask questions concerning the experiment or the
procedures involved.
8.
Know that consent to participate in the medical experiment may be withdrawn at
any time, and that the subject may discontinue participation in the medical
experiment without prejudice.
9.
Be given a copy of the signed and dated written consent form when one is
required.
10.
Be given the opportunity to decide to consent or not to consent to a medical
experiment without the intervention of any element of force, fraud, deceit,
duress, coercion or undue influence.
- 226 –Appendix A
Co-operative Research Centre for Cochlear Implant & Hearing
Aid Innovation
CONSENT FORM FOR SUBJECTS PARTICIPATING IN THE CONVENTIONAL
HEARING-AID TRIAL
Additional details re research study: to be read in conjunction with the Royal
Victorian Eye and Ear Hospital Experimental Subject’s Statement of Rights.
Aims of the research:
The broad aim of this research is to obtain a baseline fitting with digital behind-the-earhearing aids to subjects considering cochlear implantation. The study will compare the
new hearing aids with each participant’s own hearing-aid(s). It is expected that subjects
will either experience no change, or show an improvement in speech perception when
wearing the new hearing aids.
Description of the procedure:
Subjects will be provided and fitted with the hearing-aid(s). They will be asked to use
these aid(s) as much as possible in their normal daily lives for a period of 4-5 weeks.
Subjects will be asked to attend the research centre for weekly sessions to take part in
a formal evaluation of the experimental aids. The ability of each subject to understand
speech and other sounds in a variety of listening conditions will be evaluated. The
device(s) and batteries will be provided free of charge. Subjects will be reimbursed for
all costs related to their participation in the study, such as travelling expenses. A
qualified audiologist will perform the fitting and testing of the devices.
Possible benefits:
It is hoped that the hearing aids will provide users with improved perception of sounds,
especially speech. The results of the evaluations may enable further improvements to
be made in hearing-aid design and fitting, and therefore may also be of assistance to
other people with a hearing impairment.
Possible risks:
It is not envisaged that there would be any major risks involved in this trial other
than those associated with fitting and using hearing-aids in general. For
example, there is a small chance that subjects may find certain sounds
unpleasant and/or excessively loud during testing and/or when wearing the
hearing aid(s). Exposure to abnormally loud sounds, particularly over long
- 227 periods of time, could result in temporarily or permanently increased levels of
tinnitus, or loss of hearing sensitivity. Ear mould impressions will be taken for
certain subjects. In a small number of cases, subjects may experience
discomfort during this procedure. As subjects will be expected to wear the
experimental hearing aid(s) for most of the day, some other discomfort
problems may be experienced that are common with hearing aid use (for
example, moisture and/or wax build-up in the ear canal). Itchiness, skin
irritation, or swelling in the ear canal and/or concha can also be experienced
should the ear mould not fit optimally.
- 228 –Appendix A
Co-operative Research Centre for Cochlear Implants & Hearing
Aid Innovation
CONSENT FORM FOR SUBJECTS PARTICIPATING IN THE TESTS FOR HIGH
FREQUENCY HEARING LOSS
Additional details re research study: to be read in conjunction with the Royal Victorian
Eye and Ear Hospital Experimental Subject’s Statement of Rights.
Aims of the research:
To carry out a variety of tests to determine whether subjects would be suitable for a
transpositional hearing aid.
Description of the procedure:
A full case history will be determined for each subject. Subjects will be then be tested in
a sound proof booth. Subjects will be required to respond to sound or speech stimuli
according to the method required for each individual test. Subjects will be reimbursed
for all travelling expenses. A qualified audiologist will perform all testing.
Possible benefits:
There will be no immediate benefit to the individual subject. It is the intention of this
study to provide information that may be of benefit for future hearing aid studies
relating to frequency transposition.
Possible risks:
It is not envisaged that there would be any risk involved in the testing process. There
is a small chance that subjects may find certain sounds unpleasant and/or excessively
loud. Certain subjects may find the wearing of headphones uncomfortable for extended
periods of time.
- 229 -
Co-operative Research Centre for Cochlear Implant & Hearing
Aid Innovation
CONSENT FORM FOR SUBJECTS PARTICIPATING IN THE EXPERIMENTAL
HEARING-AID TRIAL
Additional details re research study: to be read in conjunction with the Royal Victorian
Eye and Ear Hospital Experimental Subject’s Statement of Rights.
Aims of the research:
The broad aim of this research is to find better ways of processing sounds so that
users of hearing aids can understand more speech. The research will be divided into
two parts. The first part will compare the experimental device with each participant’s
own hearing-aid(s). The experimental device is a new, high-powered, behind-the-ear
hearing aid. The second part of the study will compare this device with a frequencytransposing hearing-aid.
Description of the procedure:
Subjects will be provided and fitted with the experimental hearing-aid(s). They will be
asked to use these aid(s) as much as possible in their normal daily lives for a period of
several weeks. Subjects will be asked to attend the research centre for weekly
sessions to take part in a formal evaluation of the experimental aids. The ability of each
subject to understand speech and other sounds in a variety of listening conditions will
be evaluated. The device(s) and batteries will be provided free of charge. Subjects will
be reimbursed for all costs related to their participation in the study, such as travelling
expenses. A qualified audiologist will perform the fitting and testing of the devices.
Later, suitable subjects may be fitted with an experimental frequency-transposing
hearing aid. A similar procedure to that described above will be followed to evaluate the
performance of that device.
Possible benefits:
It is hoped that the experimental devices will provide users with improved perception of
sounds, especially speech. The results of the evaluations may enable further
improvements to be made in hearing-aid design and fitting, and therefore may also be
of assistance to other people with a hearing impairment.
- 230 –Appendix A
Possible risks:
It is not envisaged that there would be any major risks involved in this trial other than
those associated with fitting and using hearing-aids in general. For example, there is a
small chance that subjects may find certain sounds unpleasant and/or excessively loud
during testing and/or when wearing the hearing aid(s). Exposure to abnormally loud
sounds, particularly over long periods of time, could result in temporarily or
permanently increased levels of tinnitus, or loss of hearing sensitivity. Ear mould
impressions will be taken for certain subjects. In a small number of cases, subjects
may experience discomfort during this procedure. As subjects will be expected to wear
the experimental hearing aid(s) for most of the day, some other discomfort problems
may be experienced that are common with hearing aid use (for example, moisture
and/or wax build-up in the ear canal). Itchiness, skin irritation, or swelling in the ear
canal and/or concha can also be experienced should the ear mould not fit optimally.
- 231 -
1. PARTICIPANT INFORMATION AND CONSENT FORM
Version 4
Date: 19th May 2005
Site Melbourne
Full Project Title: ELECTRIC STIMULATION OF THE COCHLEA WITH
RESIDUAL ACOUSTIC HEARING
Principal Researcher: Hugh McDermott
Associate Researcher(s): Richard Dowell, Robert Briggs, Andrea Simpson,
Catherine Sucher, Liz Winton
This Participant Information and Consent Form is 8 pages long. Please make
sure you have all the pages.
1. Your Consent
You are invited to take part in this research project.
This Participant Information contains detailed information about the research
project. Its purpose is to explain to you as openly and clearly as possible all the
procedures involved in this project before you decide whether or not to take part
in it.
Please read this Participant Information carefully. Feel free to ask questions
about any information in the document. You may also wish to discuss the
project with a relative or friend or your local health worker. Feel free to do this.
Once you understand what the project is about and if you agree to take part in
it, you will be asked to sign the Consent Form. By signing the Consent Form,
you indicate that you understand the information and that you give your consent
to participate in the research project.
You will be given a copy of the Participant Information and Consent Form to
- 232 –Appendix A
keep as a record.
2. Purpose and Background
The purpose of this project is:
•
Aims
To evaluate the effect of providing electric hearing (via a standard cochlear
implant) in regions of profound high-frequency hearing loss together with
acoustic hearing (via a hearing-aid) in regions of normal to moderately impaired
low-frequency hearing. Subjects participating in this research will include adults
with a severe high-frequency hearing impairment, but who have relatively good
hearing sensitivity in the low frequencies. It is expected that subjects will
typically find that conventional hearing-aids are of little benefit in providing highfrequency information. This study will determine whether speech understanding
can be enhanced by means of the cochlear implant while maintaining
thresholds of acoustic hearing similar to those measured preoperatively.
•
Background
Many people with a moderate to severe hearing loss do not obtain adequate
performance from current hearing aids. In particular, people who have a severe
loss in the high frequencies, but who have good hearing sensitivity in the low
frequencies, often find conventional hearing aids of little use. Currently they are
not viewed as candidates for conventional cochlear implantation since their
residual low-frequency hearing and relatively good speech perception abilities,
compared to conventional cochlear implant candidates, precludes them.
Provision of high-frequency speech information, even if made audible, often
provides no benefit and may sometimes result in decreased speech perception
abilities.
Use of a cochlear implant, involving non-traumatic insertion of a standard
electrode array, may enable additional auditory cues to be provided to such
people with minimal risk of damage to their residual hearing. Electrical
stimulation of the cochlea could provide acoustic high-frequency cues that
would otherwise be inaudible. Provision of these auditory cues is expected to
result in better perception of speech and other sounds.
- 233 •
Previous experience
An initial study at the University of Iowa has demonstrated that a short electrode
array can be placed within the cochlea while maintaining useful low-frequency
hearing. All of the patients implanted thus far show improved speech perception
abilities with the joint use of the cochlear implant and a hearing aid in the
implanted ear.
•
You are invited to participate in this research project because
The configuration of your hearing loss indicates that you have would have
extreme difficulty obtaining high-frequency auditory cues from any conventional
hearing aids, regardless of how they are fitted.
•
Total number of people who will participate in this project
10-20
3. Procedures
Participation in this project will involve:
Subjects will be asked to participate in a number of sessions lasting
approximately 1-2 hours each. They will listen to a variety of sound stimuli
(including speech) presented through either loudspeakers, earphones, hearing
aids, or the cochlear implant’s sound processor(s). Subjects will be asked to
identify, compare, and/or describe each sound, with their responses then being
recorded by the tester. If required, they may be asked brief questions to clarify
or expand their responses, and to find out more about their listening
experiences.
Subjects’ hearing-aid and/or cochlear implant programs may be changed during
testing sessions. Subjects may be asked to use new programs when convenient
for periods of up to several weeks in order to become accustomed to different
signal processing conditions. For some subjects, speech perception ability may
decrease while they are wearing experimental devices, while for others it may
remain the same or improve.
- 234 –Appendix A
Preoperative assessment:
The preoperative assessment will be composed of two evaluations. The first
evaluation will be to determine if the candidate meets audiological criteria.
Audiological assessment will include measurement of audiometric thresholds,
speech discrimination assessment, and appropriate sound quality judgement
tasks.
Subjects will be required to report on relevant medical and surgical history in
order to confirm that they are medically suitable for the cochlear implantation
procedure. Information to be collected includes: the subject’s general medical
history, medications, radiological information (i.e., x-rays), otologic history, and
otologic surgical history.
The second evaluation will be to fit new hearing aids and establish baseline
speech perception measures. Candidates will be assessed with their current
amplification to evaluate their appropriateness for entrance into the study. Upon
entry, the subjects will be fitted, binaurally, with state-of-the art digital hearing
aids, which will be worn for a trial period and re-evaluated to confirm
continuance with the study.
Postoperative assessment:
Subjects will be implanted with a standard electrode array. Following surgical
implantation of the device and an adequate healing period, the implant will be
activated (usually 2-to-3 weeks after surgery) and auditory performance will be
reassessed.
Initially several visits to the clinic will be required to activate the cochlear
implant, and thereafter to optimise the programming of the hearing aid and the
implant’s sound processor. Auditory function will be reassessed postoperatively
using a selection of speech perception and sound-detection measures. Hearingaid alone, cochlear implant alone and joint-listening (i.e., hearing aid and
cochlear implant together) modes will be compared. Auditory tests may include
identification of speech signals (e.g. sentences, words, consonants, vowels,
etc., both in quiet conditions and with competing noise), discrimination of
signals (e.g. comparing the pitch, loudness, or other qualities of acoustic
sounds and electric stimuli), detection of signals (e.g. measuring hearing
- 235 thresholds for each mode of stimulation including the combined acoustic and
electric mode), localization of sound sources, perception of musical sounds, and
self-report measures (e.g. questionnaires and sound-quality judgements).
4. Additional information regarding study procedures (if required, see
appendix 1)
No additional information is required, as this study does not involve collection of
tissue samples for research purposes, genetic tests, HIV testing, or
performance of autopsy.
5. Possible Benefits
Possible benefits include:
•
It is hoped that subjects will obtain improved perception of speech and
other sounds with the combination of the cochlear implant and hearing
aid in the implanted ear, compared with the performance they received
with the hearing aid(s) alone.
•
The research also has the potential to benefit users of cochlear implants
and hearing aids in general, if improvements can be made to the way
sounds are processed.
6. Possible Risks
Possible risks, side effects, and discomforts include:
The main risks with the project are those risks associated with surgery to the
inner ear. These include the normal risks associated with general anaesthetic,
as well as other risks such as facial paralysis, dizziness, meningitis,
postoperative discomfort, and skin flap complications. There is a risk of loss of
residual hearing in the implanted ear with cochlear implantation.
Subjects can expect to experience some initial discomfort in the initial weeks
following surgery as recovery times vary among individuals.
There is a chance that subjects will be inadvertently exposed to excessively
loud or uncomfortable sounds. Exposure to abnormally intense sounds,
particularly over long periods of time, could result in temporarily or permanently
increased levels of tinnitus (noises in the ears), loss of hearing sensitivity, or
other problems. However, the risks of the experimental procedures to be used
are minimal, because the levels and duration of the test signals will be limited to
a comfortable range.
- 236 –Appendix A
It is possible that, during the course of trials of novel sound-processing
strategies, the speech perception abilities of some subjects will be reduced.
However, it is expected that any such reductions would be temporary.
There may be additional unforeseen or unknown risks associated with the
procedure.
Subjects are encouraged to suspend or end their participation in the project if
they experience any distress.
7. Other Treatments Whilst on Study
It is important to tell your doctor and the research staff about any treatments or
medications you may be taking, including non-prescription medications,
vitamins or herbal remedies and any changes to these during your participation
in the study.
8. Alternatives to Participation
Alternative procedures/alternative treatments include:
•
Subjects will be fitted with conventional hearing aids before the cochlear
implantation procedure is considered. These devices can provide useful lowand mid-frequency cues, although it is unlikely that traditional amplification
would be able to provide high-frequency cues for those individuals who are
candidates for the study. Subjects are also encouraged to explore other
options, such as assistive listening devices (e.g. headphones for TV
viewing) or attending communication information sessions (e.g. lip-reading
classes).
9. Privacy, Confidentiality and Disclosure of Information
Any information obtained in connection with this research project that can
identify you will remain confidential and will only be used for the purpose of this
research project. It will only be disclosed with your permission, except as
required by law. If you give us your permission by signing the Consent Form,
we plan to publish the results with an appropriate scientific or medical journal. In
any publication, information will be provided in such a way that you cannot be
identified. Your name will not appear on any publication, nor will any of your
personal details be given out without first obtaining your permission. It is
desirable that your family doctor be advised of your decision to participate in
this research project. By signing the Consent Form, you agree to your family
doctor being notified of your decision to participate in this research project. Your
- 237 health records and any information obtained during the study are subject to
inspection (for the purpose of verifying the procedures and the data) by the
Food and Drug Administration (FDA) of the United States of America (USA),
other national drug regulatory authorities such as the Australian Government’s
Therapeutic Goods Administration (TGA), or as required by law. By signing the
attached Consent Form, you authorise release of, or access to, this confidential
information to the relevant study personnel and regulatory authorities as noted
above.
10. New Information Arising During the Project
During the research project, new information about the risks and benefits of the
project may become known to the researchers. If this occurs, you will be told
about this new information. This new information may mean that you can no
longer participate in this research. If this occurs, the person(s) supervising the
research will stop your participation.
In all cases, you will be offered all available care to suit your needs and medical
condition.
11. Results of Project
It is usual for a number of years to pass before definitive results of this type of
study are available. These are published in medical journals which are available
to the public. You should feel free to ask your doctor about this.
12. Further Information or Any Problems
If you require further information or if you have any problems concerning this
project (for example, any side effects), you can contact the principal researcher,
or Richard Dowell, Robert Briggs, Andrea Simpson, Catherine Sucher, or Liz
Winton. The researchers responsible for this project are:
Hugh McDermott
(03) 9929 8665
Richard Dowell
(03) 9667 7548
Robert Briggs
(03) 9416 4133
Andrea Simpson
(03) 9662 1414
Catherine Sucher
(03) 9929 8661
- 238 –Appendix A
Liz Winton
(03) 9929 8019
13. Other Issues
If you have any complaints about any aspect of the project, the way it is being
conducted or any questions about your rights as a research participant, then
you may contact:
Secretary, Human Research and Ethics Committee
Ph:
(03) 9929 8525
You will need to tell the Secretary the name of one of the researchers given in
section 12 above.
14. Participation is Voluntary
Participation in any research project is voluntary. If you do not wish to take part
you are not obliged to. If you decide to take part and later change your mind,
you are free to withdraw from the project at any stage. Your decision whether to
take part or not to take part, or to take part and then withdraw, will not affect
your routine treatment, your relationship with those treating you or your
relationship with the Eye and Ear Hospital.
Before you make your decision, a member of the research team will be
available so that you can ask any questions you have about the research
project. You can ask for any information you want. Sign the Consent Form only
after you have had a chance to ask your questions and have received
satisfactory answers.
If you decide to withdraw from this project, please notify a member of the
research team before you withdraw. This notice will allow that person or the
research supervisor to inform you if there are any health risks or special
requirements linked to withdrawing.
15. Reimbursement for your travel costs
You will not be paid for your participation in this trial. However, you will be
reimbursed for any travel costs that you incur as a result of participating in this
trial (e.g. $30 for taxi fares each visit).
16. Ethical Guidelines
- 239 This project will be carried out according to the National Statement on Ethical
Conduct in Research Involving Humans (June 1999) produced by the National
Health and Medical Research Council of Australia. This statement has been
developed to protect the interests of people who agree to participate in human
research studies.
The ethical aspects of this research project have been approved by the Human
Research Ethics Committee of the Royal Victorian Eye & Ear Hospital.
17. Injury
In the event that you suffer an injury as a result of participating in this research
project, hospital care and treatment will be provided by the public health service
at no extra cost to you.
18. Termination of the Study
This research project may be stopped for a variety of reasons. These may
include reasons such as unacceptable side effects.
- 240 –Appendix A
Participant Consent Form
Version 3
Date
Site: Melbourne
Full Project Title: ELECTRIC STIMULATION OF THE COCHLEA WITH
RESIDUAL ACOUSTIC HEARING
Principal Researcher: Hugh McDermott
Associate Researchers: Richard Dowell, Robert Briggs, Andrea Simpson,
Catherine Sucher, Liz Winton
•
•
•
•
•
I have read, or have had read to me in my first language, and I
understand the Participant Information version 3 dated 7 April 2005.
I have had an opportunity to ask questions and I am satisfied with the
answers I have received.
I freely agree to participate in this project according to the conditions in
the Participant Information.
I have a copy of the Participant Information and Consent Form to keep.
I understand that the researcher has agreed not to reveal my identity and
personal details if information about this project is published or presented
in any public form.
Participant’s Name (printed) ……………………………………………………
Signature………………………..
Date………………..
Name of Witness to Participant’s Signature (printed) ……………………………
Signature ………………………..
Date ………………..
- 241 -
Declaration by researcher*: I have given a verbal explanation of the research
project, its procedures and risks and I believe that the participant has
understood that explanation.
Researcher’s Name (printed) ……………………………………………………
Signature………………………..
Date…………………
- 242 –Appendix A
- 243 -
APPENDIX B
1. Simpson, A., Hersbach, A.A., McDermott, H.J., 2005a. Improvements
in speech perception with an experimental nonlinear frequency-compression
hearing device. Int J Audiol., 44, 281-292.
2. Simpson, A., McDermott, H.J., Dowell, R.C., 2005c. Benefits of
audibility for listeners with severe high-frequency hearing loss. Hearing
Research, 210, 42-52.
3. Simpson, A., Hersbach, A.A., McDermott, H.J., 2006. Frequency
compression outcomes for listeners with steeply sloping audiograms. Int J
Audiol., 45, 619 - 629.
- 244 –Appendix B
Downloaded By: [University Of Melbourne] At: 01:27 30 April 2007
Original Article
International Journal of Audiology 2005; 44:281 /292
Andrea Simpson*,$
Adam A. Hersbach*,$
Hugh J. McDermott*,%
*The Cooperative Research Centre for
Cochlear Implant and Hearing Aid
Innovation, $The Bionic Ear Institute,
and %Department of Otolaryngology,
The University of Melbourne,
East Melbourne, Australia
Improvements in speech perception with an
experimental nonlinear frequency
compression hearing device
Mejorı́a en la percepción del lenguaje con un
dispositivo auditivo experimental de compresión
no lineal de la frecuencia
Key Words
Hearing impairment
Hearing aid users
Frequency compression
Abstract
Sumario
The performance of an experimental frequency compression hearing device was evaluated using tests of speech
understanding in quiet. The device compressed frequencies above a programmable cut-off, resulting in those
parts of the input signal being shifted to lower frequencies. Below the cut-off, signals were amplified without
frequency shifting. Subjects were experienced hearing aid
users with moderate-to-severe sensorineural hearing loss
and sloping audiograms. Their recognition of monosyllabic words was tested using the experimental device in
comparison with conventional hearing aids. Of the 17
subjects, eight showed a significant score improvement
(p B/0.05), whereas one subject showed a significant score
decrease. Some of the improvements may have resulted
from the better audibility provided in the high frequencies
by the experimental device in comparison with the
conventional aids. However, a subsequent study found
that increasing the high-frequency gain in the conventional aids did not produce equivalent perceptual benefits.
Se evaluó el desempeño de un dispositivo auditivo
experimental de compresión de la frecuencia utilizando
pruebas de comprensión del lenguaje en silencio. El
dispositivo comprimı́a las frecuencias por encima de un
corte programable, conllevando que estas partes de la
señal de ingreso se trasladaban a las frecuencias graves.
Por debajo del corte, las señales se amplificaban sin
cambio de frecuencia. Los sujetos eran usuarios
experimentados de auxiliares auditivos, portadores de
hipoacusias sensorineurales moderadas a severas, con un
perfil audiométrico descendente. El reconocimiento de
palabras monosilábicas fue evaluado usando el dispositivo experimental y comparándolo con auxiliares auditivos
convencionales. De los 17 sujetos, ocho mostraron una
mejorı́a significativa en la puntuación (p B/0.05), mientras
que un sujeto mostró un descenso significativo. Algunos
de los buenos resultados pueden deberse a la mejor
audibilidad resultante en las frecuencias agudas con el
dispositivo experimental, comparado con los auxiliares
convencionales. Sin embargo, un estudio subsecuente
mostró que el incremento de la ganancia en las frecuencias agudas de los auxiliares auditivos convencionales no
produjo beneficios perceptuales equivalentes.
Many hearing-impaired individuals have a greater loss of
hearing sensitivity at high frequencies than at low frequencies.
Although conventional amplification can provide usable lowfrequency information, the amplified high-frequency sounds of
speech are often inaudible due to the severity and configuration
of the hearing loss. In addition, there is evidence which
suggests that individuals with severe high-frequency hearing
loss may not obtain speech perception benefits even if these
frequencies are made audible (Ching et al, 1998; Hogan &
Turner, 1998; Murray & Byrne, 1986; Rankovic, 1991).
Clinicians are recommended to use caution in providing
amplification in the high frequencies when the hearing loss
in this region is greater than approximately 55 dB HL (Baer
et al, 2002; Hogan & Turner, 1998; Vickers et al, 2001). There
is a need, therefore, to consider alternative sound-processing
options for individuals with severe high-frequency losses.
Ideally this processing would preserve useful low-frequency
amplification and provide some additional high-frequency
information.
Various sound-processing schemes have been developed over
the past decades that have attempted to present information
from high-frequency regions of speech at lower frequencies.
The main objective of these schemes has been to utilize the
listener’s low-frequency hearing while providing additional highfrequency information. In extreme cases, frequency lowering
(also known as ‘frequency shifting’ or ‘transposition’) may be
the only way of providing this extra information acoustically.
Some early attempts at frequency shifting converted signals in
the 3 /6 kHz region into low-frequency noise below 1.5 kHz, by
passing signals through a nonlinear modulator (Johansson,
1961; Wedenberg, 1961). No significant improvements in speech
intelligibility were found with the device (Ling, 1968; Velmans &
Marcuson, 1983). One shortcoming of the scheme was that the
processing did not allow for the preservation of significant
details of the spectral shape of the incoming signals.
Alternative schemes have been developed which did present
some information about high-frequency spectral shape. In one
such scheme, Velmans (1974) separated signals into low-pass and
ISSN 1499-2027 print/ISSN 1708-8186 online
DOI: 10.1080/14992020500060636
# 2005 British Society of Audiology, International
Society of Audiology, and Nordic Audiological Society
Andrea Simpson
The Cooperative Research Centre for Cochlear Implant and Hearing Aid
Innovation, 384 /388 Albert Street, East Melbourne 3002, Australia.
E-mail: [email protected]
Received:
June 11, 2004
Accepted:
November 3, 2004
Downloaded By: [University Of Melbourne] At: 01:27 30 April 2007
high-pass bands, with a crossover frequency of 4 kHz. A
constant value of 4 kHz was subtracted from each frequency
present in the high-pass band and the resulting signals were
mixed with those obtained from the low-pass band. Although
some positive results were reported for this scheme for both
intelligibility and articulation of speech (Velmans, 1973, 1975),
the processing may have had some disadvantages. For example,
some perceptual information may have been lost, as the
frequency ratios in the high-frequency band were not preserved
when shifted to lower frequencies. It is possible that the scheme
may have provided some additional high-frequency information
at the expense of other perceptual cues by overlapping the
shifted and unshifted signals.
Another experimental processing scheme that preserved
aspects of the spectral shape analyzed the high-frequency region
of sounds using a bank of band-pass filters (Posen et al, 1993).
The envelopes of signals in these filters were estimated and used
to modulate the amplitudes of an equal number of signal
generators, which produced either pure tones or narrow-band
noises at frequencies lower than those of the corresponding
filters. The unmodified low-frequency signals were combined
with the outputs of the signal generators. This scheme was
reported to provide speech intelligibility benefits, but only after
listeners underwent a period of auditory training.
Proportional frequency shifting, using a ‘slow play’ method, is
an alternative sound-processing technique that is currently
available commercially (Bennett & Byers, 1967; McDermott et
al, 1999; McDermott & Knight, 2001). Segments of the speech
signal are recorded and then played back at a slower speed than
employed for recording. The TranSonic and ImpaCt DSR675
(AVR Communications Ltd.) are two commercially available
hearing instruments that incorporate such processing. Incoming
signals dominated by components at frequencies above 2.5 kHz
are shifted down by a factor that is programmable for each
listener. If the input signal is not dominated by frequencies above
2.5 kHz, then signals are amplified with no frequency shifting.
Positive outcomes were reported when the TranSonic was fitted
to a small number of hearing-impaired children (Davis-Penn &
Ross, 1993; Rosenhouse, 1990). A similar study with adults
found that two of the four subjects tested demonstrated speech
perception benefits with the device (Parent et al, 1997).
McDermott et al (1999) reported a study in which five subjects
obtained higher scores with the TranSonic than with their own
hearing aids, but suggested that the amplification characteristics
of the TranSonic in the low frequencies, rather than its frequency
shifting characteristics, may have provided most of the benefit.
Only two subjects appeared to obtain additional speech information specifically from the high-frequency signal components
after they were lowered. A more recent study (McDermott &
Knight, 2001) of the AVR ImpaCt hearing instrument found
little difference in performance between the ImpaCt aid and the
subjects’ own conventional aids. In addition, the subjects’
understanding of sentences in competing noise was significantly
poorer with the ImpaCt.
However, some listeners have obtained speech perception
benefits when listening to proportional frequency compression
(Turner & Hurtig, 1999). An advantage of this method is that
frequency ratios are preserved. In other words, the relationship
between the frequencies of different formant peaks in speech
remain constant. These ratios may be particularly important
cues for the recognition of vowels in speech (Neary, 1989).
Generally, studies investigating the perceptual performance
resulting from the use of frequency lowering schemes have
provided mixed results. Many novel processing schemes are
possible with the arrival of digital technology and may be of
benefit if appropriate fitting parameters can be found. The study
described below reports speech perception results with an
experimental frequency compression scheme. The scheme aims
to provide high-frequency speech cues while preserving the lowand mid-frequency components of speech. Experiment (a)
compared the speech understanding abilities of 17 hearingimpaired listeners when listening via the experimental device
with the frequency compression function enabled and disabled.
Subsequently, experiment (b) investigated whether the addition
of high-frequency audibility without frequency shifting would
result in similar outcomes.
Figure 1 shows schematically the hardware of the frequency
compression hearing instrument, which consists of two main
parts: a pair of modified behind-the-ear (BTE) conventional
hearing devices, and a SHARP programmable body-worn speech
processor (Zakis et al, 2001). Sound entered the system through
the microphone (or telecoil or direct audio input) of each BTE,
and was processed by the conventional hearing device. The BTEs
were modified so that their outputs were directed to the SHARP
processor. The SHARP processor then manipulated the signals to
perform frequency compression and sent the outputs to the
earphone receivers located in the BTEs. There was one cable
connecting each BTE to the SHARP processor. Typically, the
hearing aid wearer would fasten the SHARP processor to a belt or
pocket, running the cables to the BTE devices worn on each ear.
Left Receiver
Left Microphone
Left BTE
Low-pass filter
∑
Frequency
compression
processing
Right BTE
Right Microphone
Low-pass filter
∑
Right Receiver
Sharp body-worn processor
Figure 1. Block diagram of the binaural signal processing implemented in the frequency compression hearing device.
282
International Journal of Audiology, Volume 44 Number 5
Fin ;
Fout 1p
FLPF
Finp ;
Fin BFLPF
Fin ]FLPF
where
Fin /input frequency
Fout /output frequency
Improvements in speech perception with an
experimental nonlinear frequency compression hearing device
FLPF /cutoff frequency
p /compression exponent
An example of this relation is shown in Figure 2, where the cutoff frequency (FLPF ) is 1.6 kHz, and the frequency compression
exponent (p ) is 0.5. In addition to the frequency compression,
amplitude compression was applied to signals in the high-pass
band for all subjects. As discussed later, an amplitude compression ratio of at least 2:1 was selected to improve audibility and
potentially intelligibility of speech, while avoiding possible
deleterious effects on signal quality (Plomp, 1994). The amplitude compression had an attack time of 5 ms and a release time
of 30 ms with the compression threshold set to approximately
25 dB SPL.
Method
Subjects
Seventeen hearing-impaired adults, comprising 7 women and 10
men, participated in the study. All subjects had participated in
related research projects at the time of testing. Relevant
information about the subjects is provided in Table 1. The
majority of subjects had moderate-to-severe hearing losses. Their
hearing threshold levels, measured conventionally under headphones, are listed in Table 2.
For all subjects, hearing losses were assumed to have primarily
a sensorineural origin, based on standard air and bone conduction audiometry. Five of the subjects had one unaidable ear, in
which cases, the fitting and evaluation of the hearing aids was
carried out only on the other ear (see Table 1). In these cases,
only one BTE and audio signal processing channel of the
experimental device was utilized. All subjects were experienced
hearing aid users. Subjects were not paid for their participation
in the experiments, although expenses such as travel costs were
reimbursed. The use of human subjects in these experiments was
in accordance with the ‘Guiding principles for research involving
human or animal subjects,’ and was approved by the Human
Research Ethics Committee of the Royal Victorian Eye and Ear
Hospital, Melbourne, Australia.
10
5
Output Frequency (kHz)
Downloaded By: [University Of Melbourne] At: 01:27 30 April 2007
As illustrated in Figure 1, low-frequency signals were processed separately for each ear in the modified BTE hearing aids,
whereas one frequency compressed signal was derived from the
left input only, and delivered to both left and right receivers.
Audio signals from each input were passed to the SHARP
processor, and sampled at 14.4 kHz after being band-limited by
an anti-aliasing filter. The samples were then converted into the
digital domain. The input signals were separated into a low-pass
band and a high-pass band. The cross-over frequency between
the two bands, called the cut-off frequency, was programmable
to facilitate appropriate fitting of the device to each user. The
low-pass band was created by means of a finite impulse response
(FIR) filter having 127 taps. This filter had a typical slope above
the cut-off frequency of /35 dB/octave. Each of the signals from
the left and right BTE hearing aids was processed by a separate
FIR filter, but with identical characteristics. This enabled the
low-pass band from each BTE device to be returned separately
to that device without those signals undergoing any additional
processing. To generate the frequency compressed part of the
output signal, the input from only the left BTE was processed
further. Samples of this signal were processed in overlapping
blocks, each comprising 256 samples. Each block was windowed
in the time domain using the product of a Hamming and a sinc
function (Crochiere & Rabiner, 1983). The data blocks were then
folded into 128-point sequences for processing by a fast Fourier
transform (FFT). The FFT was executed after each set of 32
input samples had been collected (i.e. at intervals of approximately 2.2 ms). The real and imaginary components of the FFT
output in the frequency domain were processed to extract
magnitude and phase information. The phase angles were used
to generate estimates of the time rate of change of phase at each
FFT bin frequency. This processing, which is similar to the
technique known as the phase vocoder (Moore, 1990) provides
an estimate of the magnitude and instantaneous frequency of the
input signal for each FFT component. To create the output
signal for the frequency compression hearing aid, a subset of
these estimates was selected and used to control the amplitude
and frequency of a corresponding bank of sine oscillators. The
first 24 FFT bins above the chosen cut-off frequency were
selected and assigned to 24 oscillators. The amplitude data from
the FFT were modified to control the output level at each
frequency according to the needs of the aid user. Frequency
compression was achieved by modifying the frequency estimates
from the FFT before using them to control the frequencies of the
corresponding oscillators. As described below, higher frequencies were shifted downwards to a larger extent than lower
frequencies. The outputs of the 24 sine oscillators were then
summed. Finally, this composite signal was summed with the
low-band signal from the left BTE input to generate the left
earphone output, and with the low-band signal from the right
BTE input to generate the right earphone output.
The general form of the relation between the output and the
input frequency was:
Cut-Off Frequency
2
1
0.5
0.2
0.1
0.1
0.2
0.5
1
2
5
10
Input Frequency (kHz)
Figure 2. An example of the input frequency to output
frequency relationship implemented in the frequency compression hearing device.
Simpson/Hersbach/McDermott
283
Downloaded By: [University Of Melbourne] At: 01:27 30 April 2007
Table 1. Relevant information about the subjects, their own hearing aids, and the conventional device used in the study
Subject
Age
(yrs)
Probable etiology
of hearing loss
S02
S08
72
79
F
M
S10
47
F
S11
S12
S13
S14
54
59
69
71
F
F
M
M
S15
69
M
S16
54
F
Ototoxic drugs, family
history
Resound Canta 7
S17
63
M
Industrial noise exposure
Widex Senso CX/
S18
S19
74
M
F
Industrial noise exposure
Viral infection
Bernafon AA310
Phonak Claro
21 daz
S23
S24
S25
80
69
74
M
M
M
Industrial noise exposure
Industrial noise exposure
Unknown
Starkey Sequel
Bernafon AA310
Bernafon LS16D
S26
S28
64
57
M
F
Industrial noise exposure
Unknown
Phonak PPCLC
Bernafon PB675
Sex
Unknown
Industrial noise
exposure
Unknown
Premature presbyacusis
Otosclerosis
Industrial noise exposure
Industrial noise
exposure, cholesteatoma
Unknown
Type of own
hearing aids
Features of own
hearing aids
Processing strategy
of conventional
hearing device
Ears
fitted
Bernafon RB15
Bernafon RB15
Digitally programmable
Digitally programmable
Linear
Linear
Left
Binaural
Canal Aid Dynamic
Equalizer II
Phonak PICS
Starkey A-13
Phonak PICS
Bernafon RB15
Fully automatic digital
aid
Digitally programmable
Analog
Digitally programmable
Digitally programmable
Linear
Binaural
Compression
Linear
Linear
Compression
Binaural
Right
Binaural
Left
Widex Senso CX/
Fully automatic digital
aid with adaptive
directional microphone
Fully automatic digital
aid with adaptive
directional microphone
Fully automatic digital
aid with adaptive
directional microphone
Digitally programmable
Fully automatic digital
aid with adaptive
directional microphone
Analog
Digitally programmable
Digital hearing aid with
adaptive directional
microphone
Analog
Digitally programmable
Compression
Binaural
Linear
Binaural
Compression
Binaural
Compression
Linear
Binaural
Right
Compression
Compression
Compression
Binaural
Binaural
Binaural
Linear
Linear
Binaural
Left
Experiment (a)
The hearing aid usage and medical history of each subject was
documented during the first test session. A pure-tone audiogram, including both air and bone conduction, was obtained,
and the electro-acoustic characteristics of each subject’s own
hearing aids were measured and recorded. Table 1 includes
relevant details of each subject’s own aids.
Each subject was fitted with identical conventional hearing
aids. These were behind-the-ear (BTE) digital power instruments
(Phonak Supero 412). They were specified to have a maximum
output and a maximum gain of approximately 140 dB SPL and
80 dB, respectively (measured in an ear simulator), and were
designed to be most suitable for people with hearing threshold
levels that exceed 50 dB HL at all frequencies. The gain and
amplitude compression characteristics were separately adjustable
in five partially-overlapping frequency bands. The amplitude
compression attack and release times were 5 and 30 ms,
respectively.
The conventional hearing instruments were fitted to each
subject using appropriate fitting software (Phonak Fitting
Guideline version 8.1), with which user-selectable normal and
noise-reduction programs were created. The subjects’ pure-tone
thresholds were entered into the fitting software to derive an
initial fitting suggestion based on the manufacturer’s recommendation. When necessary, these settings were altered at the
follow-up sessions based on subject feedback. As recommended
by the fitting software program, linear amplification was selected
when the average hearing loss at 0.5, 1, 2, and 3 kHz was equal
to or greater than 70 dB HL, whereas amplitude compression
was selected in cases where the average hearing thresholds were
284
International Journal of Audiology, Volume 44 Number 5
STIMULI
FOR DEVICE FITTING
As described later, equal loudness level measurements were used
to fit the frequency compression device. Third-octave narrowband noises with duration of 0.5 s were selected. Each noise
stimulus was separated from the next stimulus by a silent interval
of duration 0.5 s. The noises were ramped in level at each end
with linear ramps of duration 30 ms. The level of each noise was
set to equal the average third-octave level at the corresponding
frequency of the international long-term average speech spectrum with an overall level of 70 dB SPL (Byrne et al, 1994).
DEVICE
FITTING METHOD
Downloaded By: [University Of Melbourne] At: 01:27 30 April 2007
Table 2. Hearing threshold levels (dB HL) and selected cut-off frequencies for the subjects who participated in the study
Frequency (kHz)
Subject
Ear
0.25
0.5
0.75
1
1.5
2
3
4
8
S02
S08
L
L
R
L
R
L
R
R
L
R
L
L
R
L
R
L
R
L
R
R
L
R
L
R
L
R
L
R
L
35
50
45
45
40
50
50
70
50
30
40
50
40
60
70
55
55
40
30
55
40
35
20
30
55
55
55
55
90
50
65
65
65
60
60
55
65
35
35
40
75
50
75
95
50
50
55
45
60
50
45
35
45
60
60
70
70
90
70
80
80
85
85
90
80
80
90
70
55
60
60
45
85
75
105
105
55
60
65
55
65
55
55
60
60
70
65
65
70
105
85
120*
115
90
95
90
75
70
80
65
35
95
75
115
115
65
70
80
65
70
60
55
60
75
75
75
75
70
105
85
120*
120
90
90
90
80
85
90
70
50
85
70
120
120
70
75
80
90
80
70
65
65
80
90
70
75
75
105
95
120*
115
100
95
85
105
95
90
85
105
90
65
120*
120*
80
85
70
85
80
85
80
80
105
120
80
100
90
105
120
120*
115
120
100
90
95
95
100
90
100
110
75
120*
120*
85
110
80
80
85
105
100
75
85
120
95
120
110
120
110*
110*
110*
110*
110*
75
110*
110*
90
75
110*
110*
75
110*
110*
110*
110*
70
70
110*
110*
110*
100
100
110*
110*
110*
110*
110*
S10
S11
S12
S13
S14
S15
S16
S17
S18
S19
S23
S24
S25
S26
S28
80
75
60
90
95
45
50
65
70
100
Selected cut-off
frequency (kHz)
2
1.6
2.5
2
2.5
2.5
2.5
2
1
2.5
2
2.5
2.5
2
2
2
1.6
Note: Asterisks indicate levels that were limited by the maximum output of the audiometer.
lower (better) than 70 dB HL. Eight of the 17 subjects were fitted
with amplitude compression programs. Table 1 provides relevant
details of the final programs selected for each subject.
Speech perception comparisons were made between this
hearing device and the subjects’ own hearing aids. Open-set
testing was carried out using monosyllabic word lists at a level of
65 dBA. Subjects were selected for the current study only if their
speech perception results with the conventional hearing device
were approximately the same as, or better than, those obtained
with their own hearing aids. Each subject had been wearing the
conventional hearing devices for several months prior to the
commencement of the current study.
To fit the experimental frequency compression scheme, each
subject’s fitting parameters determined as described above were
programmed into a modified conventional hearing device. A cutoff frequency was chosen individually for each subject based on
the audiogram. As a general rule, the cut-off was initially set to
the lowest frequency at which conventional amplification provided the listener with inadequate audibility. For the purposes of
this study, hearing thresholds of 90 dB HL or greater were
assumed to prevent the listener from obtaining adequate
audibility. For example, a cut-off frequency of 2 kHz was chosen
if the subject’s hearing threshold level was more than 90 dB HL
for frequencies above 2 kHz. The initial cut-off frequency was
adjusted by the clinician if the subject was dissatisfied with the
sound quality of the device. Table 2 provides the final cut-off
frequencies selected for each subject. The frequency compression
exponent of 0.5 was kept constant for all subjects. In addition,
amplitude compression was enabled for frequencies above the
cut-off frequency for all subjects. For subjects whose conventional device fittings were linear, the amplitude compression ratio
was set to 2:1. For subjects fitted with amplitude compression in
the conventional device, the amplitude compression ratio was
increased to 2:1, if the amplitude compression ratio in the
conventional device was lower than 2:1. The amplitude compression ratio was not altered if it was greater than 2:1. The amplitude
compression characteristics for frequencies below the cut-off
frequency were identical to those of the conventional device for
each subject. That is, nine subjects retained linear fittings in the
low frequencies, whilst eight subjects retained amplitude compression fittings in the low frequencies.
For each subject, the loudness of the test signals specified
above was approximately equalized across frequency. To achieve
this, amplification of the high frequencies was adjusted by means
of one of two methods: loudness balancing or loudness rating.
The loudness balancing method was the primary fitting method.
Improvements in speech perception with an
experimental nonlinear frequency compression hearing device
Simpson/Hersbach/McDermott
285
Downloaded By: [University Of Melbourne] At: 01:27 30 April 2007
If the subject had difficulty with the task (e.g. due to confusion
between pitch and loudness), the loudness rating method was
used instead.
In the loudness balancing method, a reference frequency was
chosen for each subject below the cut-off frequency. A reference
frequency of 1 kHz was appropriate for all subjects except for
S16. A lower reference frequency of 0.8 kHz was selected for this
subject.
For all testing in the sound-field, each subject was seated at a
distance of 1 m directly in front of the loudspeaker. The subject
was asked to compare the loudness of a pair of narrow-band
noises, one at the reference frequency, and the other at a selected
frequency in the compressed signal (test frequency). The order in
which the signals were presented in each pair was randomized.
Amplification of the test frequency was adjusted manually by
the clinician until the test noise was perceived by the subject to
have the same loudness as the reference. Other frequencies within
the compressed signal were then adjusted in the same way.
Across subjects, the most commonly tested frequencies
were 2, 2.5, 3.15, and 4 kHz. These frequencies were presented
in a random order. Finally, the signals at frequencies within the
compressed signal were compared in loudness with each other,
and further adjustments were made when necessary.
In the loudness rating method, subjects were given a
categorical list of nine loudness levels. The subject was asked
to indicate the loudness of the noise stimuli by pointing to a
category on the printed list. The nine categories were: very soft,
soft, comfortable but slightly soft, comfortable, comfortable but
slightly loud, loud but OK, uncomfortably loud, extremely
uncomfortable, and painfully loud. Narrow-band noises were
presented at 1.6, 2, 2.5, 3.15, 4, and 5 kHz. The amplification of
each frequency was adjusted manually by the clinician for each
narrow-band noise until the subject reported each noise to be
‘comfortable but slightly soft’.
After fitting, subjects were asked to wear the device away from
the laboratory and provide feedback about it at two follow-up
sessions. If necessary, the parameters of the shifted signal, such
as the loudness level and the cut-off frequency, were adjusted at
these sessions. No further adjustments were made to the
program for the remainder of the trial.
SPEECH
TEST STIMULI
A test of word recognition was carried out in quiet. Consonant /
vowel nucleus / consonant (CNC) monosyllabic word lists were
presented from audio recordings (Peterson & Lehiste, 1962).
There were 50 words per list, spoken by a woman with an
average Australian accent. No lists of words (other than practice
lists) were repeated for any subject during the trial. The order in
which lists were presented to subjects across sessions was
randomized. The average level of the words, when measured at
the subject’s listening position (about 1 m from the loudspeaker)
was 55 / 60 dBA. These levels, which are similar to the levels of
speech in normal conversation, were generally perceived as
comfortably loud when heard by the subjects through their
hearing aids.
SPEECH
device was set for comfortable listening to speech at a conversational level in quiet conditions. For most subjects, this was the
default volume control setting. This setting was noted and fixed
for all following test sessions. A practice CNC word list was then
presented to familiarize subjects with the testing procedure and
materials. Subjects were instructed to repeat each word immediately after hearing it, and to guess if unsure. Responses from
the practice list were excluded from the data analysis. After the
practice list, four lists were used to test subjects in each of the
two conditions: (1) using the conventional hearing device; and
(2) using the experimental frequency compression scheme.
Subjects’ responses were analyzed to determine the number of
phonemes correctly recognized out of a total of 150 phonemes
per list.
A counterbalanced sequence of testing was applied in an
attempt to minimize the confounding effects of acclimatization
over time. Initially, subjects were tested with two lists using their
conventional hearing aids. They were then asked to take the
experimental device home, and use it in place of their conventional hearing aids as much as possible. Each subject wore the
experimental device for a total period of 4 / 6 weeks. The four
CNC word tests were carried out during the final two sessions of
this period. At the end of the trial period, subjects reverted to
wearing the conventional hearing aids. After a further two
weeks, a final test was carried out to obtain scores for two
additional CNC word lists using the conventional hearing aids.
The four sets of scores for each of the two conditions were
averaged and compared.
Experiment (b)
EXPERIMENTAL
AID FITTING
It was anticipated that speech perception differences obtained
with the frequency compression scheme could result from
amplification differences (and therefore audibility differences)
in the high frequencies with the experimental device in comparison with the conventionally fitted instruments. To investigate
this possibility, further testing was carried out with a subset of
subjects who obtained significant improvements when using the
experimental hearing aid.
The experimental processor was re-programmed with the
frequency compression factor set to 1. This resulted in no
shifting of the high-frequency input signal. The high-frequency
signal was adjusted according to the loudness rating or loudness
balancing fitting method used in experiment (a), so that narrow
band noise stimuli were equalized in loudness across frequency.
The new program was referred to as ‘added high-frequency gain’
(HFG).
No other changes were made to the program unless feedback
oscillation occurred. In such cases, the gains of the high
frequencies were reduced until oscillation stopped. It was
necessary to reduce gains for the majority of subjects included
in the experiment. The experimental processor’s output was then
measured to compare the final gains with those of the original
conventional hearing instrument program. The subject was
asked to wear the HFG program for two weeks.
TEST METHOD
For all evaluations of speech intelligibility, each subject was
tested individually in a medium sized sound-attenuating booth.
The volume control on each subject’s conventional hearing
SPEECH
286
International Journal of Audiology, Volume 44 Number 5
TEST METHOD
Speech understanding was assessed with most subjects who
obtained score increases with the experimental device using a
Relative Sensation Level (dB)
(a)
FrC
20
10
0
–10
1
1.2
1.6
2
2.5
3.15
4
5
Frequency (kHz)
40
Relative Sensation Level (dB)
Downloaded By: [University Of Melbourne] At: 01:27 30 April 2007
the hearing aid output and hearing threshold at the same input
frequency. For the frequency compression device, the sensation
level is reported at each frequency for the same input range;
however, the sensation level was calculated by taking the
difference between the hearing aid output level and hearing
threshold at the shifted output frequency. For binaurally fitted
subjects, the maximum sensation level of the two ears at each
frequency was plotted. In order to compare sensation levels
across subjects, relative sensation levels were calculated by
normalizing the values against the sensation levels produced
by the conventional device for each subject.
Panel (a) of Figure 3 shows the sensation levels relative to the
conventional device for the frequency compression device
averaged across all 17 subjects. The panel shows that the
frequency compression device provided a higher sensation level
than the conventional device for frequencies above 2 kHz.
Relative sensation levels for the subset of 5 subjects who were
fitted with the HFG program in experiment (b) are shown in
panel (b), and discussed later.
30
(b)
30
Experiment (a)
FrC
HFG
20
10
0
–10
–20
1
1.2
1.6
2
2.5
3.15
4
5
Frequency (kHz)
Figure 3. High frequency sensation levels relative to those for
the conventional device. Panel (a) shows the relative sensation
levels for the frequency compression device averaged across all
17 subjects. Panel (b) shows relative sensation levels for the
frequency compression and HFG programs for the subset of 5
subjects who were fitted with the HFG program in experiment
(b). Error bars indicate plus and minus one standard deviation
from the mean.
further four CNC word lists. These lists were tested over two test
sessions. The four sets of scores were averaged and compared
with those from the previous two test conditions.
Results
Figure 3 shows the relative sensation levels across frequency for
the frequency compression device and the HFG program
averaged across subjects. Output levels in dB SPL were measured
via a Brűel & Kjær Type 4157 ear simulator for a speech-shaped
input at 70 dB SPL for the conventional device, the frequency
compression device, and the HFG program. The sensation level
at each input frequency is the level in dB by which the output
signal of the hearing aid exceeds the subjects’ hearing threshold.
For the conventional device and the HFG program, the
sensation level reported at each frequency was calculated from
Improvements in speech perception with an
experimental nonlinear frequency compression hearing device
For the CNC word tests, the mean phoneme scores obtained by
each subject with their conventional hearing devices and with the
experimental device are shown in Figure 4. Panel (a) shows the
percent of all phonemes (vowels and consonants) in the
monosyllabic words recognized correctly by each subject when
using each of the two types of hearing aid. Panels (b), (c), and (d)
show the percent correctly recognized of consonants, consonants
containing frication (i.e. /f/, /s/, /sh/, /v/, /z/, /th/, /ch/, and /dj/),
and vowels, respectively, extracted from the same data. (In the
following, the above subset of consonants will be referred to as
fricatives, for the sake of brevity, although the last two are
usually classified as affricates.)
Across the 17 subjects, speech recognition scores were
compared with the two hearing aid processing schemes by
means of a two-factor analysis of variance. A statistically
significant improvement for the frequency compression scheme
(FrC) over the conventional hearing device (CD) was found
(p B/0.001) for phoneme, consonant, fricative, and vowel scores.
However, there was also a significant interaction term (subject /
scheme, pB/0.001), confirming that individual subjects performed with the different schemes in different ways. Therefore,
each subject’s data were analysed separately with pair-wise
comparisons using the Holm-Sidak test (Hochberg & Tamhane,
1987). For each subject, and for each type of score (phonemes,
consonants, fricatives, and vowels), statistical significance levels
are indicated in Figure 4 by asterisks at the bottom of each
panel.
Of the 17 subjects, eight (S10, S12, S13, S15, S17, S25, S26,
S28) obtained a significant (p B/0.05) phoneme score increase
with the experimental transposition scheme, eight subjects (S02,
S08, S11, S14, S16, S19, S23, S24) showed no significant change
in scores, and one subject (S18) showed a significant score
decrease with the experimental scheme.
Experiment (b)
Of the eight subjects (S10, S12, S13, S15, S17, S25, S26, S28)
who performed better with the experimental scheme, only five
subjects (S12, S13, S15, S25, and S26) participated in this trial.
S17 was unavailable at the scheduled time of testing. Subjects
Simpson/Hersbach/McDermott
287
Phonemes correctly identified (%)
(b )
90
CD
FrC
80
80
70
70
60
60
50
50
40
40
30
30
20
20
10
10
0
0
S02 S08 S10 S11 S12 S13 S14 S15 S16 S17 S18 S19 S23 S24 S25 S26 S28 Mean
***
100
Fricatives correctly identified (%)
90
** ***
**
* **
S02 S08 S10 S11 S12 S13 S14 S15 S16 S17 S18 S19 S23 S24 S25 S26 S28 Mean
* *** *** ***
*
*** **
**
**
* *** *** ***
100
(d)
(c )
90
80
80
70
70
60
60
50
50
40
40
30
30
20
20
10
10
0
Vowels correctly identified (%)
Downloaded By: [University Of Melbourne] At: 01:27 30 April 2007
90
(a )
Consonants correctly identified (%)
100
100
0
S02 S08 S10 S11 S12 S13 S14 S15 S16 S17 S18 S19 S23 S24 S25 S26 S28 Mean
** ** * ***
***
***
* ** *** ***
S02 S08 S10 S11 S12 S13 S14 S15 S16 S17 S18 S19 S23 S24 S25 S26 S28 Mean
***
***
*
*
* ***
Figure 4. Mean phoneme [panel (a)], consonant (b), fricative (c), and vowel (d) scores obtained by the 17 hearing-impaired subjects
when listening to monosyllabic words. Unfilled columns show scores obtained using the conventional hearing devices (CD), and filled
columns show scores obtained using the experimental frequency-compression scheme (FrC). Scores averaged across subjects are
shown in the pair of rightmost columns in each panel, with error bars indicating one standard deviation. Statistical significance is
shown by asterisk symbols: *0.01 B/pB/0.05, **0.001 B/pB/0.01, ***pB/0.001.
S10 and S28 were unable to wear the device due to
feedback oscillation occurring. To avoid oscillation, the
high-frequency gain was reduced to a setting that was within
2 dB of that for the standard conventional hearing device fitting
at each frequency. The resulting fitting and the standard
conventional hearing device fitting were considered to be
too similar for further testing to provide any additional
information.
Some feedback occurred with the additional high-frequency
gain program at the target gain settings for the remaining five
subjects (S12, S13, S15, S25, and S26). The high-frequency gain
was reduced as required until feedback stopped. These subjects
were asked to wear the processor away from the laboratory, as
the new gain settings were higher than those of the original
conventional hearing device fitting.
Panel (b) of Figure 3 shows the average sensation levels
relative to those for the conventional device for the frequency
compression and HFG programs for the five subjects who
were successfully fitted with the HFG program. The HFG
program did provide a higher sensation level, on average, at
frequencies above 2 kHz, than the conventional device. However,
due to feedback oscillation occurring, the sensation levels
attained were lower than with the frequency compression
device.
The mean phoneme scores obtained by each subject for the
conventional hearing device (CD), the experimental device
(FrC), and the additional high-frequency gain program (HFG)
are shown in Figure 5. For each subject and for each type of
score (phonemes, consonants, fricatives, and vowels), statistical
significance levels comparing the test conditions are indicated on
each graph by asterisks at the bottom. A two-factor analysis of
variance was carried out on all subjects’ phoneme, consonant,
fricative, and vowel scores. Across all subjects, there was no
statistically significant difference between the added HFG
program and the conventional hearing device. Speech recognition scores for individual subjects were better with the
added HFG program than with the original conventional
hearing device for S13 and S25, but the score increase for all
phonemes was significant for S13 only. However, the frequency
compression program resulted in the greatest increase in
phoneme scores. Subjects S12 and S15 showed a significant
decrease in fricative and phoneme scores, respectively, with the
high-frequency gain program relative to the original conventional hearing device.
288
International Journal of Audiology, Volume 44 Number 5
100
(a )
CD
FrC
HFG
80
80
70
70
60
60
50
50
40
40
30
30
20
20
10
10
0
0
S12
CD vs. FrC
***
CD vs. HFG
FrC vs. HFG
S 13
S15
***
**
*
*
***
S25
***
S26
Mean
S12
S13
S15
S 25
S26
Mean
***
***
***
***
***
*
* **
***
*
***
*
***
***
***
***
***
***
100
90
110
(c )
(d )
100
90
80
80
70
70
60
60
50
50
40
40
30
30
20
20
10
10
0
Vowels correctly identified (%)
Phonemes correctly identified (%)
90
Consonants correctly identified (%)
(b )
90
Fricatives correctly identified (%)
Downloaded By: [University Of Melbourne] At: 01:27 30 April 2007
100
0
S12
S13
S15
CD vs. FrC **
CD vs. HFG *
***
**
***
FrC vs. HFG ***
**
***
S25
*
*
S26
Mean
***
***
***
***
S1 2
S13
S15
***
**
S 25
S26
*
*
**
Mean
***
***
Figure 5. Mean phoneme [panel (a)], consonant (b), fricative (c), and vowel (d) scores obtained by the 5 subjects who participated in
the experiment with the experimental processor set to have additional high frequency gain (HFG) with no frequency shifting. Subjects
listened to monosyllabic words. Unfilled columns show scores obtained with the conventional hearing device (CD), filled columns
show scores obtained with the experimental frequency-compression processor (FrC), and hatched columns show scores obtained
with the additional high frequency gain program (HFG). Scores averaged across subjects are shown in the group of rightmost
columns, with error bars indicating one standard deviation. Statistical significance is shown by asterisk symbols: *0.01 B/pB/0.05,
**0.001 B/pB/0.01, ***p B/0.001.
Discussion
For the 17 subjects who participated in the study, the frequency
compression scheme provided perceptual performance superior,
on average, to the performance of the conventionally fitted
hearing aids for words presented at a moderate level in quiet
conditions. Furthermore, there is evidence that the positive
outcome observed for this group of subjects could not have been
achieved by increasing the high-frequency audibility of the
hearing aid alone.
For all subjects who participated in Experiment (b), it was not
possible to increase high-frequency audibility substantially without activating the frequency compression function. This was
particularly the case for S10 and S28, for whom no increase
Improvements in speech perception with an
experimental nonlinear frequency compression hearing device
in gain above the standard conventional hearing device fitting
could be achieved without feedback oscillation occurring.
For these subjects, it is difficult to separate the effects of
frequency compression from those of increased audibility,
and it is likely that both these factors resulted in the increase
in speech recognition scores they obtained. The compression
of selected high-frequency signals into a narrower frequency
range, where the listener had better residual hearing, could
have enabled more effective use of the additional audible
speech information. These are encouraging results for those
hearing impairments where conventional hearing instruments
are limited in their ability to provide audibility in the high
frequencies.
Simpson/Hersbach/McDermott
289
Manner
Place
Voicing
Sonorance
Frication
P
B
T
D
K
G
F
V
TH
S
Z
SH
H
M
N
W
Y
L
R
CH
DJ
0
0
0
0
0
0
0
1
0
0
0
1
0
0
0
0
1
1
0
0
0
2
0
0
0
0
2
1
0
0
1
3
0
0
1
1
3
1
0
1
1
4
0
0
1
1
1
0
0
1
1
1
1
0
1
1
5
0
0
1
1
6
0
0
0
2
0
1
1
0
2
1
1
1
0
3
0
1
0
0
3
5
1
0
0
3
1
1
0
0
3
1
1
0
0
4
5
0
0
1
4
5
1
0
1
The experimental scheme does not preserve frequency ratios
for those high frequencies that are compressed. It seems unlikely
that speech perception would be affected greatly by this, as all of
the first-formant frequency range, and most of the secondformant frequency range, are below the cut-off frequency for
the majority of subjects. Low- and mid-frequency information
appear to remain preserved with the frequency compression
device, as vowel scores for the group were either improved
or unaffected. It is a possibility, however, that perception of
other sounds, such as music, may be affected adversely. It is
important to extend testing with a variety of sound stimuli,
including music and background noise, before generalizations
can be made.
The improvement in fricative scores for this group of subjects
provides some evidence that the experimental scheme provided
additional high-frequency speech cues. Although the average
score for fricative identification remained low for the group
(see Figure 4), frequency compression appeared to provide
some beneficial speech cues for identifying this group of
consonants.
Confusion matrices were constructed from the subjects’ CNC
word responses with both the conventional hearing aids as well
as with the experimental hearing devices. Sequential Information
Analysis (SINFA) was used to analyse these matrices to
determine how much information about the consonant articulatory features of manner, place, voicing, sonorance, and
frication were conveyed in both conditions (Miller & Nicely,
100
Relative Information Transmitted (%)
Downloaded By: [University Of Melbourne] At: 01:27 30 April 2007
Table 3. Feature allocation used in the Sequential Information Transmission Analyses
1955; Wang & Bilger, 1973). Table 3 shows the phoneme feature
allocation used in the analysis. The information transmitted
from stimulus to response was computed using a custom
software program. The results are summarized in Figure 6.
Panel (a) of Figure 6 shows the average information transmission results for those subjects (S02, S08, S11, S14, S16, S18, S19,
S23, S24) who did not show an improvement in speech
perception with the frequency compression device. Not surprisingly, a one-way ANOVA found no significant difference in the
relative information transmitted for any articulatory feature
between the conventional device and experimental device for this
group of subjects.
Panel (b) of Figure 6 shows the average information transmission results for those subjects (S10, S12, S13, S15, S17, S25, S26,
S28) who showed an improvement in speech perception with the
frequency compression device. A one-way ANOVA found that
these subjects obtained substantially more information about the
consonant features of place (p /0.02, t/2.613) and frication
(p /0.02, t/2.552) with the frequency compression device.
Information about these features is conveyed largely by highfrequency speech components (Sher & Owens, 1974). The
increase in frication and place information seems to indicate
that the experimental scheme provided these subjects with
additional information, particularly about high frequencies. In
contrast, features that rely more on low-frequency speech
components, such as sonorance, were unaffected by the
frequency compression scheme.
100
(a)
90
CD
FrC
80
80
70
70
60
60
50
50
40
40
30
30
20
20
10
10
0
(b)
90
0
Manner
Place
Voicing
Sonorance
Frication
Manner
Place
Voicing
Sonorance
*
Frication
*
Figure 6. Results of information transmission analyses for subjects’ CNC word responses. Unfilled columns show scores obtained
using the conventional hearing devices (CD), and filled columns show scores obtained using the experimental frequency compression
scheme (FrC). Panel (a) on the left shows average results for subjects who showed no significant phoneme score differences with the
experimental scheme. Panel (b) on the right shows the average results for those subjects who obtained significant score increases with
the experimental scheme. Statistically significant score differences (p B/0.05) are shown by asterisk symbols.
290
International Journal of Audiology, Volume 44 Number 5
Downloaded By: [University Of Melbourne] At: 01:27 30 April 2007
However, the experimental scheme may not suit all listeners.
One subject (S18) did show a significant decrease in speech
perception scores with the frequency compression scheme
(see Figure 4). It is possible that different fitting parameters
might have resulted in a more positive outcome. It is important
to refine the procedures used to fit frequency-shifting devices
for suitable hearing-impaired listeners. This should include
investigation of alternative frequency-shifting schemes and
selection of the signal-processing parameters in those schemes,
including the frequency compression exponent and the cut-off
frequency.
Finally, it should be noted that, in the experimental device,
there was one frequency compressed signal, which was presented
to both ears. The experimental scheme did not allow for cases
where subjects have asymmetrical hearing thresholds in the high
frequencies. This is not an ideal fitting solution, and adjustments
will need to be made to the processing scheme in future to
accommodate binaural hearing threshold differences.
Conclusions
In summary, experiment (a) evaluated the performance of a
frequency compression device by comparing the speech understanding abilities of 17 hearing-impaired listeners when the
frequency compression function was enabled and disabled.
Experiment (b) investigated whether the addition of highfrequency audibility without frequency shifting could produce
similar outcomes for those subjects who obtained significantly
higher speech scores with the frequency compression device than
with their conventionally fitted device.
The results of these evaluations of the frequency compression
scheme can be summarized as follows.
1. Use of the experimental hearing aid provided better recognition of monosyllabic words than conventionally fitted
hearing aids, on average, for 17 adult subjects.
2. Of the 17 subjects, eight obtained a significant phoneme
score increase with the experimental scheme, eight showed
no significant change in scores, and one showed a significant
score decrease in comparison with conventional hearing aids.
3. For most subjects, increasing high-frequency audibility via
the hearing instrument alone, without frequency compression, was not possible without feedback occurring. It is
difficult to separate the effects of frequency shifting from
those of increased audibility, and it is likely that both these
factors resulted in the improvements in speech understanding
found with the frequency compression scheme.
4. For all subjects, greater high-frequency sensation levels were
achieved by enabling frequency compression than by a
conventional hearing aid fitting.
5. Information transmission analyses of the subjects’ CNC
word responses showed that subjects who obtained significant score increases with the experimental scheme obtained
more information about the high-frequency consonant
features of place and frication in particular.
Acknowledgements
The authors are grateful for the financial support of the
Commonwealth of Australia through the Cooperative Research
Improvements in speech perception with an
experimental nonlinear frequency compression hearing device
Centres program. Phonak Hearing Systems provided the conventional hearing instruments used in the study. The third
author received financial support from the Garnett Passe and
Rodney Williams Memorial Foundation. The comments of two
anonymous reviewers and the Associate Editor (Dr Brian
Moore) on a previous version of the manuscript are greatly
appreciated. We are thankful for the contributions of many
colleagues to this work, particularly Cathy Sucher, Rodney
Millard, Justin Zakis, and Dr Wai Kong Lai. We are also
especially grateful to all of the hearing aid users who participated
as subjects in the experiments.
References
Baer, T., Moore, B.C. & Kluk, K. 2002. Effects of low pass filtering on
the intelligibility of speech in noise for people with and without dead
regions at high frequencies. J Acoust Soc Am , 112, 1133 /44.
Bennett, D.S. & Byers, V.W. 1967. Increased intelligibility in the
hypacusic by slow-play frequency transposition. Journal of Auditory
Research , 7, 107 /118.
Byrne, D., Dillon, H., Tran, K., Arlinger, S., Wilbraham, K., et al. 1994.
An international comparison of long-term average speech spectra. J
Acoust Soc Am , 96, 2108 /2120.
Ching, T.Y., Dillon, H. & Byrne, D. 1998. Speech recognition of hearingimpaired listeners: predictions from audibility and the limited role of
high-frequency amplification. J Acoust Soc Am , 103, 1128 /40.
Crochiere, R.E. & Rabiner, L.R. 1983. Multirate Digital Signal Processing . Englewood Cliffs, NJ: Prentice-Hall.
Davis-Penn, W. & Ross, M. 1993. Paediatric experiences with frequency
transposing. Hearing Instruments, 44, 26 /32.
Hochberg, Y. & Tamhane, A.C. 1987. Multiple Comparison Procedures.
Hobeken, NJ: Wiley.
Hogan, C.A. & Turner, C.W. 1998. High-frequency audibility: benefits
for hearing-impaired listeners. J Acoust Soc Am , 104, 432 /41.
Johansson, B. 1961. A new coding amplifier system for the severely hard
of hearing. Proceedings of the 3rd International Congress on
Acoustics Stuttgart 1959 , 2, 655 /657.
Ling, D. 1968. Three experiments on frequency transposition. American
Annals of the Deaf , 113, 283 /294.
McDermott, H.J., Dorkos, V.P., Dean, M.R. & Ching, T.Y. 1999.
Improvements in speech perception with use of the AVR TranSonic
frequency-transposing hearing aid. J Speech Lang Hear Res, 42,
1323 /35.
McDermott, H.J. & Knight, M.R. 2001. Preliminary results with the
AVR ImpaCt frequency-transposing hearing aid. J Am Acad Audiol ,
12, 121 /7.
Miller, G. & Nicely, P. 1955. An analysis of perceptual confusions among
some English consonants. J Acoust Soc Am , 27, 623 /637.
Moore, F.R. 1990. Elements of Computer Music . Englewood Cliffs, NJ:
Prentice-Hall.
Murray, N.B. & Byrne, D. 1986. Performance of hearing-impaired and
normal hearing listeners with various high-frequency cut-offs in
hearing aids. Australian Journal of Audiology, 8, 21 /28.
Neary, T. 1989. Static, dynamic, and relational properties in vowel
perception. J Acoust Soc Am , 85, 2088 /2113.
Parent, T.C., Chmiel, R. & Jerger, J. 1997. Comparison of performance
with frequency transposition hearing aids and conventional hearing
aids. J Am Acad Audiol , 9, 67 /77.
Peterson, G. & Lehiste, I. 1962. Revised CNC lists for auditory tests. J
Speech Hear Dis, 27, 62 /70.
Plomp, R. 1994. Noise, amplification, and compression: considerations
of three main issues in hearing aid design. Ear Hear, 15, 2 /12.
Posen, M.P., Reed, C.M. & Braida, L.D. 1993. Intelligibility of
frequency-lowered speech produced by a channel vocoder. J Rehabil
Res Dev, 30, 26 /38.
Rankovic, C.M. 1991. An application of the articulation index to hearing
aid fitting. J Speech Hear Res, 34, 391 /402.
Rosenhouse, J. 1990. A new transposition device for the deaf. The
Hearing Journal , 43, 20 /25.
Simpson/Hersbach/McDermott
291
Downloaded By: [University Of Melbourne] At: 01:27 30 April 2007
Sher, A.E. & Owens, E. 1974. Consonant confusions associated with
hearing loss above 2000 Hz. J Speech Hear Res, 17, 669 /81.
Turner, C.W. & Hurtig, R. R. 1999. Proportional frequency compression
of speech for listeners with sensorineural hearing loss. J Acoust Soc
Am , 106, 877 /86.
Velmans, M. 1973. Speech imitation in simulated deafness, using visual
cues and ‘recoded’ auditory information. Language and Speech , 16,
224 /36.
Velmans, M. 1974. The design of speech recoding devices for the deaf. Br
J Audiol , 8, 1 /5.
Velmans, M. 1975. Effects of frequency ‘recoding’ on the articulation
learning of perceptively deaf children. Language and Speech , 18,
180 /93.
Velmans, M. & Marcuson, M. 1983. The acceptability of spectrumpreserving and spectrum-destroying transposition to severely hearing-impaired listeners. Br J Audiol , 17, 17 /26.
Vickers, D.A., Moore, B.C. & Baer, T. 2001. Effects of low-pass
filtering on the intelligibility of speech in quiet for people with and
without dead regions at high frequencies. J Acoust Soc Am , 110,
1164 /75.
Wang, M.D. & Bilger, R.C. 1973. Consonant confusions in noise: a study
of perceptual features. J Acoust Soc Am , 54, 1248 /66.
Wedenberg, E. 1961. Auditory training of the severely hard of hearing
using a coding amplifier. Proceedings of the 3rd International
Congress on Acoustics Stuttgart 1959 , 2, 658 /660.
Zakis J.A., McDermott H.J. & Fisher M. 2001. A high performance
digital hearing aid for advanced sound processing research. Proceedings of ‘‘Biomedical Research in 2001’’, IEEE Engineering in
Medicine and Biology Society (Victorian Chapter) , Monash University, Victoria, Australia, 17 /20.
292
International Journal of Audiology, Volume 44 Number 5
Hearing Research 210 (2005) 42–52
www.elsevier.com/locate/heares
Benefits of audibility for listeners with severe
high-frequency hearing loss
Andrea Simpson
a,b,c,*
, Hugh J. McDermott
a,c
, Richard C. Dowell
c
a
c
The Cooperative Research Centre for Cochlear Implant and Hearing Aid Innovation, The University of Melbourne,
384–388 Albert Street, East Melbourne 3002, Australia
b
The Bionic Ear Institute, The University of Melbourne, 384–388 Albert Street, East Melbourne 3002, Australia
Department of Otolaryngology, The University of Melbourne, 384–388 Albert Street, East Melbourne 3002, Australia
Received 25 May 2005; accepted 8 July 2005
Available online 30 August 2005
Abstract
A consonant identification test was carried out with 10 hearing-impaired listeners under various low-pass filter conditions. Subjects were also tested for cochlear dead regions with the TEN test. All subjects had moderate-to-severe high-frequency hearing
losses. Consonant recognition was tested under conditions in which the speech signals were highly audible to subjects for frequencies
up to the low-pass filter cut-off. Extensive dead regions were found for one subject with the TEN test. The remaining subjects may
have had dead regions above 3 kHz, because of the severity of their hearing losses, but these could not be demonstrated with the
TEN test. Average consonant scores for the subject group improved significantly (p < 0.05) with increasing audibility of high-frequency components of the speech signal. There were no cases of speech perception being reduced with increasing bandwidth. Nine of
the subjects showed improvements in scores with increasing audibility, whereas the remaining subject showed little change in scores.
For this subject, speech perception results were consistent with the TEN test findings. In general, the results suggest that listeners
with severe high-frequency losses are often able to make some use of high-frequency speech cues if these cues can be made audible.
2005 Elsevier B.V. All rights reserved.
Keywords: High-frequency hearing impairment; Hearing aids; Cochlear dead regions
1. Introduction
One of the primary effects of hearing impairment is a
decrease in the audibility of sounds due to threshold elevaAbbreviations: ACT, air conduction threshold; AI, articulation
index; ANOVA, analysis of variance; C, conversion factor; CD, compact disc; CRC, cooperative research centre; dB, decibel; dBA, decibel
sound pressure level, A-weighted; ERB, equivalent rectangular bandwidth; f, frequency; G, gain; HL, hearing level; Hz, hertz; LSD, least
significant difference; ms, millisecond; NAL, National Acoustic Laboratories; p, probability; PTC, psychophysical tuning curve; s, second;
SPL, sound pressure level; TEN, threshold equalizing noise; VCV,
vowel–consonant–vowel; VU, volume unit; 4IFC, four interval forced
choice
*
Corresponding author. Tel.: +61 3 9662 1414; fax: +61 3 9663
6086.
E-mail address: [email protected] (A. Simpson).
0378-5955/$ - see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.heares.2005.07.001
tion. Individuals with severe losses may not hear certain
speech sounds unless they are presented at high intensity
levels. Providing high-frequency audibility to listeners
with a severe high-frequency hearing loss presents several
challenges for the clinician when fitting a hearing aid.
Firstly, the level difference between discomfort and the
threshold of audibility (i.e. the dynamic range of hearing)
is usually much less than that of a normally hearing ear
(Steinberg and Gardner, 1937). Secondly, acoustic feedback will often limit the amount of usable gain, dependent
on individual characteristics, such as how effectively the
ear canal is sealed by the earmold.
In addition, there is evidence to suggest that the typical hearing aidÕs upper frequency limit may not be high
enough to ensure the audibility of certain high-frequency
fricatives, such as /s/. For example, Stelmachowicz et al.
A. Simpson et al. / Hearing Research 210 (2005) 42–52
(2001) evaluated fricative perception for a group of hearing impaired subjects under various low-pass filter conditions. To obtain optimum performance for female and
child speakers, it was found that the bandwidth was required to be at least 9 kHz.
Research indicates that some individuals with moderate to severe hearing impairment do not benefit from
amplification at high frequencies (Hogan and Turner,
1998; Murray and Byrne, 1986; Rankovic, 1991). Murray and Byrne (1986) fitted hearing aids to 5 subjects
with sloping high-frequency hearing losses. The amplification was shaped to suit their hearing losses but had
variations in the high-frequency cut-off selected. Three
of the subjects obtained the highest speech recognition
scores with bandwidths below the widest bandwidth
available. Of the two remaining subjects, one showed a
small improvement in speech scores with the highest
bandwidth. The fifth subjectÕs scores were highly variable and inconclusive. Murray and Byrne (1986)
concluded that, in some conditions, stimulation with
high-frequency sounds may produce distortion in the
auditory system.
A similar study by Rankovic (1991) compared two
different amplification strategies to a procedure that attempted to maximize audibility of speech (AImax). Due
to limitations of the equipment, maximum audibility
was not achieved for any subject, although the AImax
strategy did result in higher audibility than the other
two strategies. Of the 13 ears tested, there were no cases
where AImax significantly improved performance over
both alternative strategies. The majority showed no decrease in performance with the AImax condition. Four
ears performed more poorly with AImax when compared
to conditions where audibility was less than the maximum possible. All of these ears had relatively normal
low-frequency hearing thresholds and sloping high-frequency hearing losses. Even though AImax generally provided the highest gain, it did not provide the highest
speech recognition score. These studies seemed to indicate that the widest bandwidth, or greatest audibility,
is not always desirable in an amplification system.
Hogan and Turner (1998) investigated the benefits of
providing high-frequency information to hearing impaired listeners. Nonsense syllables were low-pass filtered at various cut-off frequencies and presented to
both normally hearing and hearing impaired listeners
at various levels via headphones. It is possible that, even
at the highest presentation level, some components of
the speech signals remained inaudible to some subjects.
The articulation index (AI) was used to quantify audibility for each condition and for each listener. Listeners
with mild high-frequency loss performed similarly to
the normally hearing group with scores increasing as
audibility increased. Recognition scores of those subjects with severe losses, however, showed that increasing
the audibility resulted in no further improvements in
43
scores, on average. The hearing impaired listeners were
not making use of high-frequency information above
4000 Hz in some cases where subjects had a hearing loss
of less than 55 dB HL.
The lack of benefit of high-frequency amplification,
when found, may be due to the presence of a dead region, or non-functioning inner hair cells within a certain
region of the cochlea (Moore and Glasberg, 1997;
Thornton and Abbas, 1980). The threshold-equalizing
noise (TEN) test (Moore, 2001; Moore et al., 2000a)
was developed as a means for clinicians to be able to
identify dead regions. It is based upon the detection of
sinusoids in the presence of a broadband noise. Normally hearing listeners and listeners with hearing impairment but without dead regions should show almost
equal masked thresholds (in dB SPL) over a wide frequency range with the test. Individuals with dead regions should show abnormally high masked thresholds
at one or more frequencies when the signal frequency
falls within a dead region.
Moore et al. (2000a) assessed the validity of the test
by measuring psychophysical tuning curves (PTCs) with
14 hearing impaired listeners and then performing the
TEN test with the same group of subjects. When a dead
region occurs at high frequencies, the signal is usually
detected via a downward spread of excitation, and the
tip of the PTC is shifted towards lower frequencies. A
good correspondence was found between the results obtained using the TEN test and the measured PTCs. In
particular, abnormally high masked thresholds in the
TEN test were associated with PTCs with shifted tips.
In further studies related to the TEN test, vowel–consonant–vowel (VCV) nonsense syllables were presented
to 10 subjects under various low-pass filter conditions
(Vickers et al., 2001). The limits of any dead region
for each subject were defined by both PTC measurement
and the TEN test results. It was found that consonant
identification generally improved with increasing filter
cut-off frequency for subjects with no dead regions.
For subjects with dead regions, performance improved
with increasing cut-off frequency until the cut-off frequency was somewhat above the estimated edge frequency of the dead region, but hardly changed with
further increases. A few subjects had worsening performance with further increases in the cut-off frequency,
although this worsening was statistically significant for
only one subject. The results seem to indicate that there
is little benefit in amplifying frequencies well inside a
dead region, but there may be some benefit in amplifying frequencies up to approximately one octave above
the estimated edge frequency of the dead region. However, a confounding factor in the study was that the subjects with dead regions had more severe high-frequency
hearing losses than the subjects without dead regions.
This may have resulted in insufficient audibility of the
speech stimuli for those subjects with dead regions at
44
A. Simpson et al. / Hearing Research 210 (2005) 42–52
certain frequencies. This lack of audibility was likely to
have contributed to the poor scores these subjects
obtained.
Rankovic (2002) calculated articulation index (AI)
predictions using the Fletcher method of calculation
(Fletcher, 1953; Fletcher and Galt, 1950) for the data
published by Vickers et al. (2001). The AI was generally
accurate in predicting the consonant recognition test
scores of the subjects irrespective of the presence or absence of dead regions. For subjects without dead regions, both the AI and consonant recognition score
increased with increasing cut-off frequency. For subjects
with dead regions, there was little or no AI change with
increasing low-pass filter frequency, particularly at cutoffs within the dead regions. This finding implies that
audibility was an important factor in the Vickers et al.
(2001) study. Dybala and Thibodeau (2002) agreed with
those conclusions, suggesting that listeners in the deadregion group did not benefit from high-frequency amplification because of inaudibility.
However, the AI was not accurate in predicting the
incremental benefit of amplifying frequencies well above
the estimated edge frequency of a dead region (Moore,
2002). The AI predicted that speech scores should improve by 10–15%, whereas the largest improvement
found in the Vickers et al. (2001) study was only 7%.
Testing on the same subjects was also carried out
using nonsense-syllable stimuli in noise (Baer et al.,
2002). Results were similar to those obtained by Vickers
et al. (2001) with performance improving for cut-off frequencies up to 1.5–2 times the edge frequency of the
dead region, but hardly changing with further increases.
To address the issue of audibility, a modification of the
AI (Moore and Glasberg, 1998) was used to calculate
the audibility of the speech stimuli. Baer et al. (2002)
concluded that frequency components of the stimuli
within the dead regions were at least partially audible
to subjects. The authors maintained that subjects with
dead regions do not make as effective use of audible
high-frequency speech information as subjects without
dead regions.
In summary, many studies have attempted to investigate whether high-frequency amplification benefits
listeners with severe sloping hearing losses. Unfortunately, in many of these studies it was not possible to
provide full audibility of the high-frequency signal.
Audibility remains a confounding factor when determining which hearing impaired listeners could benefit from
additional high-frequency information. The TEN test
was developed as a means of providing clinicians with
a tool to diagnose dead regions, and thereby a means
for deciding when to provide high-frequency amplification. The study described below reports speech perception results for 10 subjects with moderate-to-severe
high-frequency hearing losses under various low-pass filter conditions. The confounding factor of audibility was
reduced when measuring speech perception by maximizing audibility of the speech stimuli. The TEN test was
carried out with all subjects.
2. Materials and methods
2.1. Subjects
Ten hearing-impaired adults (S01–S10) participated
in the study. All subjects had participated in related research projects at the time of testing. Relevant information about the subjects is provided in Table 1. All
subjects had moderate-to-severe high-frequency hearing
losses. Their hearing threshold levels for the tested ear,
measured using standard audiometric methods under
headphones are listed in Table 2.
For all subjects, hearing losses were assumed to have
primarily a sensorineural origin, based on standard airand bone-conduction audiometry. All subjects were
experienced hearing aid users. At the time of the study,
they were participating in a separate hearing aid trial. As
Table 1
Relevant information about the subjects who participated in the study
Subject
Ear tested
Age (yrs)
Sex
Etiology
S01
S02
S03
S04
S05
S06
S07
S08
S09
S10
R
L
R
R
R
L
R
L
R
R
82
63
59
60
80
69
74
74
69
69
M
M
F
F
M
M
M
M
M
M
Industrial noise
Industrial noise
Otosclerosis
Viral infection
Industrial noise
Industrial noise
Unknown
Industrial noise
Industrial noise
Industrial noise
exposure
exposure
exposure
exposure
exposure
exposure
exposure
Table 2
Hearing threshold levels (dB HL) for the subjects who participated in
the study
Frequency (kHz)
S01
S02
S03
S04
S05
S06
S07
S08
S09
S10
0.25
0.5
1
1.5
2
3
4
8
35
55
70
55
35
20
55
40
30
40
45
50
65
60
45
35
60
55
35
50
65
55
55
65
55
60
65
65
60
75
65
65
70
70
55
60
75
75
65
75
65
70
85
80
65
65
70
80
70
70
75
80
95
80
80
80
80
70
85
65
80
85
95
85
100
75
95
75
90
75
85
110*
110*
110*
110*
100
110*
70
75
75
Note. Asterisks indicate levels that were limited by the maximum
output of the audiometer.
A. Simpson et al. / Hearing Research 210 (2005) 42–52
a result, all subjects were wearing the same type of digital, high-powered, behind-the-ear hearing aid, fitted
generally according to the NAL-NL1 prescription (Byrne et al., 2001). They had been wearing this aid for a
period of several months prior to the commencement
of the current study.
Subjects were not paid for their participation in the
experiments, although expenses such as travel costs were
reimbursed. Their participation in these experiments
was in accordance with the ‘‘Guiding principles for
research involving human or animal subjects’’, and
was approved by the Human Research Ethics Committee of the Royal Victorian Eye and Ear Hospital, Melbourne, Australia.
2.2. Procedure
The TEN test was carried out with each subject. In
addition, aided thresholds were measured with each subject to estimate the audibility of the speech stimuli. Finally, consonant recognition was measured under
various low-pass filter conditions.
2.3. TEN test
2.3.1. Stimuli
All stimuli used in testing were obtained from the CD
‘‘Diagnosis of Dead Regions’’ developed by Moore et al.
(2000b). The CD recording contained two channels. The
first channel had several different test signals consisting
of digitally generated pure tones at the following frequencies: 250, 500, 1000, 1500, 2000, 3000, 4000, 5000,
6000, 8000, and 10000 Hz. Each test frequency was
assigned a unique track number on the CD.
The second channel on the CD was a spectrally
shaped broadband noise termed ‘‘threshold equalizing
noise’’ (TEN). The TEN noise is intended to provide
equal masked thresholds in dB SPL by producing a constant amount of excitation at each characteristic frequency in the mid-frequency range of 500–5000 Hz.
In cases without dead regions, the masked threshold
is approximately equal to the nominal level of the noise
specified in dB/ERB, where ERB stands for the equivalent rectangular bandwidth of the auditory filter
(Moore, 2001).
2.3.2. Procedure
Moore et al. (2000b) recommended a method for
carrying out the TEN test using a conventional audiometer. This method was adjusted slightly to enable testing
to be carried out using a Madsen Aurical audiometer.
The levels of the stimuli were adjusted on the Audiometer
Speech Module screen. The test instructions advise the
clinician to set both VU meters on the audiometer to read
6 dB prior to commencing. The Aurical software limits
45
the VU meters to a minimum of 5 dB. This 1 dB difference was taken into account when evaluating results.
The standard audiometric method recommended for
measuring pure-tone thresholds (Carhart and Jerger,
1959) was used to determine absolute thresholds in dB
SPL. Absolute thresholds were determined in the test
ear at the following frequencies: 250, 500, 1000, 1500,
2000, 3000, 4000, and 5000 Hz. The order of presenting
test frequencies was randomized across subjects and test
sessions.
Masked thresholds were determined next at each test
frequency. The same method was used as when measuring absolute threshold with one exception. When nearing threshold, 1 dB incremental steps were used instead
of 5 dB steps to obtain a more precise measurement.
Wherever possible, the noise level was increased and
the masked thresholds were measured again to obtain
measurements at more than one level/ERB. A high
masking level was required to mask high frequencies
as all subjects had severe high-frequency hearing losses.
For this reason, TEN levels of 81 and 91 dB/ERB were
selected when testing most subjects.
The order of presenting test frequencies was randomized across subjects and test sessions. The entire procedure of measuring both absolute and masked
thresholds was repeated at a following test session to obtain two sets of threshold data. These two sets of data
were averaged for later analysis.
2.4. Aided thresholds
Aided thresholds were used to obtain a measurement
of the audibility of the speech test stimuli for each subject. As mentioned, the aim was to maximize the audibility of the speech stimuli for each subject without
causing discomfort. To achieve this, aided thresholds
needed to fall below the minimum speech levels across
the frequency range. For the purposes of this study,
these levels were taken to be 18 dB below each onethird octave band level of the international long-term
average speech spectrum according to Byrne et al.
(1994). The hearing aid was connected to the fitting
software and the gain adjusted to be as close as possible
to the gain settings calculated by means of the formula
below.
Stimuli were presented to the hearing aid via direct
audio input. To achieve this, the hearing aid was fitted
with an audioshoe that disconnected the microphone
and delivered electric signals directly to the input of
the aidÕs amplifier. Prior to testing, signals were calibrated so that they were displayed in units of dB SPL
by the computer system used in the experiments.
The amount of gain necessary for aided thresholds to
fall below the speech minima in each one-third octave
band was determined for each subject using the following formula:
46
A. Simpson et al. / Hearing Research 210 (2005) 42–52
Gðf Þ ¼ ACTðf Þ þ Cðf Þ Sðf Þ;
where G was the gain in dB for a 2cc coupler, ACT was
the air conduction threshold in dB HL, C was the conversion factor from dB HL to dB SPL for a 2cc coupler
(Bentler and Pavlovic, 1989), S was the minimum onethird octave band level for speech at an overall level of
60 dB SPL (Byrne et al., 1994), and f was the frequency.
In the hearing aid, a linear amplification scheme was
selected, and the noise canceller and feedback canceller
were deactivated. A program that provided the required
gain at each frequency was saved into the hearing aid. In
some cases, particularly at very high frequencies, it was
not possible to achieve an aided threshold below the corresponding minimum one-third octave band level of the
speech signal. In these cases, the gain was set to the maximum value allowed by the fitting software. The use of
direct audio input ensured that no acoustic feedback
could occur with the hearing aid, and therefore enabled
the gain to be set to higher than normal values. All subsequent testing was performed with each subject wearing
this hearing aid fitted in accordance with their individually determined program.
2.4.1. Stimuli
One-third octave narrowband noises with duration
0.5 s were selected for measuring aided thresholds. The
noises were ramped in level at each end with linear
ramps of duration 30 ms. Initially the level of each narrowband noise was set to equal the one-third octave level at the corresponding frequency of the international
long-term average speech spectrum, according to Byrne
et al. (1994). The frequencies selected for this test were:
250, 500, 1000, 1600, 2000, 3150, and 4000 Hz.
2.4.2. Procedure
Subjects were always tested individually. The test aid
was fitted according to the subjectÕs program determined
above and attached to the audioshoe. The audioshoe
cable was connected to the computer soundcard. The
test aid was connected to the subjectÕs own ear-mold
and placed in the test ear. Prior to testing, the vent on
each subjectÕs mold was blocked by inserting a closed
plug to minimize sound leakage.
Subjects were seated at a desk facing a computer monitor. A series of four red boxes were displayed on the computer screen, numbered 1–4. The test was conducted using
a four interval forced choice (4IFC) method. Each box
was lit up in order (1–4). The subject was required to indicate which one of the 4 boxes corresponded to the sound
being presented. A signal was presented in only one of the
4 intervals, selected at random for each trial.
Each subjectÕs aided threshold at each frequency was
determined using an adaptive procedure. The initial level was presented twice by the program. There was a
4 dB reduction of the level once two correct responses
were recorded consecutively. The reduction in level continued in this way until the subject made an incorrect response. Once one incorrect response was made, the level
was increased by 4 dB. This point was automatically recorded as the first turning point. Once two turning
points were obtained, the incremental level change was
reduced to 2 dB. A further 6 turning points were obtained using steps of 2 dB. The program calculated the
average SPL over the last 6 turning points.
This procedure was repeated for each frequency. Initially one measurement was obtained at each frequency.
The measurement was repeated at a particular frequency
if the standard error of all the 8 turning points was
greater than or equal to 2.0 dB.
2.5. Speech recognition
Prior to speech recognition testing, signal delivery via
the audioshoe was calibrated to be compatible with the
MACarena Speech Tests system. MACarena Speech
Tests is a software program designed to present acoustic
stimuli and record subject responses. MACarena was
used in the current study to present each subject with
consonant stimuli.
2.5.1. Stimuli
The test materials comprised vowel–consonant–vowel
(VCV) nonsense syllables recorded by a male and a female speaker. The 16 consonants selected for the test
were: /p/, /t/, /k/, /b/, /d/, /g/, /m/, /n/, /s/, /sh/, /z/,
/f/, /v/, /ch/, /j/, and /th/. The initial vowel was always
the same as the final vowel.
The male-speaker stimuli were selected from recordings on the NAL CRC CD, disc 1 – ‘‘Speech and Noise
for Hearing Aid Evaluation’’ (National Acoustic Laboratories, 2000). They were embedded in the vowels /a/
and /i/, and spoken with a typical Australian accent.
The female-speaker stimuli were obtained from University College London (Markham and Hazan, 2002).
These stimuli were low-pass filtered at four cut-off
frequencies: 1.4, 2.0, 2.8, and 5.6 kHz. A 10th order Butterworth filter was selected with a slope of 60 dB/octave. The levels of the medial consonant tokens were
approximately equalized before filtering. The average
levels were approximately 60 dBA. The total number
of stimuli was 256 (i.e., 16 /aCa/ consonants + 16 /iCi/
consonants · 4 low-pass filter conditions · 2 speakers).
Each stimulus was repeated four times making a total
of 1024 tokens.
2.5.2. Procedure
A closed-set test of consonant recognition was performed, in which 16 buttons appeared on a computer
screen. These buttons were of the following format:
vCv, where the vowel was either /i/ or /a/, and C represented each consonant.
A. Simpson et al. / Hearing Research 210 (2005) 42–52
The stimuli were divided into 16 subtests consisting of
the 4 low-pass filter conditions · 2 speakers · 2 vowels.
Each stimulus was repeated twice in each subtest. Each
subtest was repeated 2 times to make a total of 4 repetitions. The stimuli were presented in a random order in
each subtest. The stimuli were divided into 4 blocks: female speaker with /a/ vowel, female speaker with /i/ vowel, male speaker with /a/ vowel, and male speaker with
/i/ vowel. The 4 blocks each contained 4 subtests, which
were the 4 low-pass filter conditions. One block was presented to each subject at each test session. The entire test
was completed over 4 test sessions. The order in which
these blocks were presented was randomized across subjects and test sessions. The low-pass filter conditions
were presented in the following order within each of
the four blocks: 5.6, 2.8, 2.0, 1.4, 1.4, 2.0, 2.8, and
5.6 kHz. Balancing the order of the filter conditions in
this way helped to compensate for possible learning effects during the testing procedure.
Prior to testing, the clinician showed the subject each
button on the screen and verbalized which sound corresponded to it. Once the clinician was confident the subject was proficient in the task, testing began. The
subjects were presented with the 16 consonant tokens
in the test set on the computer screen. They were instructed to identify which consonant they heard after
each stimulus was presented by pressing the corresponding button on the screen. There was no option for a response other than one of the 16 consonants, and it was
not possible to provide ‘‘no answer’’ as a response.
At the start of each test session, each subject performed the easiest subtest (i.e. male speaker, cut-off of
5.6 kHz, using the vowel /a/) to gain practice and confidence in the procedure. The results were recorded as
usual, but disregarded in the final analysis of the data.
3. Results
3.1. TEN test
Fig. 1 shows the TEN test results for all subjects.
Absolute thresholds are shown for each subject in dB
SPL across the frequency range. Measured masked
thresholds in dB SPL at 250, 500, 1000, 1500, 2000,
3000, 4000, and 5000 Hz are shown for each subject at
two masking levels/ERB. Each graph has two shaded
areas that are 10 dB in width. These areas represent
graphically the range beyond which a masked threshold
is considered to indicate a dead region.
Three of the 10 subjects (S01, S02, S06) showed dead
regions according to the TEN test criteria. In other
words, masked thresholds were elevated more than
10 dB above both the measured absolute thresholds
and the masking noise level for these subjects. S06
showed elevated masked thresholds at 2000, 3000, and
47
4000 Hz. S02 showed an elevated threshold measurement at 4000 Hz only. Masked thresholds for these
two subjects (S02, S06) showed a similar pattern of elevation at both noise levels of 81 and 91 dB/ERB. S01Õs
masked threshold at 1500 Hz was elevated when the
noise level was 91 dB/ERB. The threshold was not elevated at the lower noise level of 81 dB/ERB. No indications of dead regions were found for this subject at the
other frequencies tested.
Dead regions were not found at any frequency with
the remaining 7 subjects. One subject (S08) showed
approximately equal masked thresholds across the frequency range. According to TEN test principles, it can
be assumed that this subject had no dead regions. For
subjects S04, S05, S07, and S09, the TEN noise was
not sufficiently intense to produce 10 dB or more of
masking above 3000 Hz. For subjects S03 and S10, the
TEN noise was not sufficiently intense to produce
10 dB or more of masking above 4000 Hz. For example,
S07 had high absolute thresholds at 4000 and 5000 Hz.
Even the highest TEN level of 91 dB/ERB was not sufficient to mask these frequencies. For these subjects, it
was not possible to determine with the TEN test whether
or not dead regions were present at these high frequencies, but they probably did not have a dead region starting at 3000 Hz or below.
3.2. Aided thresholds and speech recognition
Fig. 2 shows each subjectÕs aided thresholds. The
shaded area represents a ‘‘speech banana’’. Values are
adapted from the long term average speech spectrum
(Byrne et al., 1994) for speech measured at an overall
average level of 70 dB SPL (approximately 65.4 dBA).
Speech stimuli in the current study were presented to
subjects at approximately 60 dBA. For Fig. 2, 5.4 dB
was subtracted from the speech spectrum of Byrne
et al. (1994) to provide an estimate of audibility for
speech at 60 dBA. The upper and lower boundaries of
the shaded area were defined by adding 12 dB and subtracting 18 dB from these average level values for each
one-third octave frequency band.
All subjects obtained aided thresholds that were
within or below the speech banana across the frequency
range. Generally, the frequencies of 3000 and 4000 Hz
were the least audible for subjects, with only S06 and
S08 obtaining aided thresholds near the lower edge of
the speech banana at both of these frequencies. Most
subjects showed the greatest audibility for the low frequencies, with aided thresholds for some subjects (S02,
S03, S06, S08) falling below the speech banana at 250
and 500 Hz.
Percentage scores for the consonant test across the
four low-pass filter conditions for each subject are also
shown in Fig. 2. Most subjectsÕ speech perception scores
improve with increasing bandwidth. The exception is
48
A. Simpson et al. / Hearing Research 210 (2005) 42–52
Fig. 1. TEN test results for subjects S1–S10. Absolute thresholds for each subject are represented by filled triangles across a frequency range of 250–
5000 Hz. Masked thresholds at two noise levels (filled and unfilled circles) are also shown. Error bars on each symbol represent plus and minus one
standard deviation from the mean. The two shaded areas on each graph are 10 dB in width. They represent graphically the point outside of which a
masked threshold is considered to indicate a dead region according to TEN test principles. Masking levels of 81 and 91 dB/ERB were used for all
subjects, except for S05 and S06 for whom masking levels of 71 and 81 dB/ERB were used. Possible cochlear dead regions are represented by
horizontal black bars for S01, S02, and S06.
S06, whose scores showed no improvement from the
2.0 kHz bandwidth upwards.
An analysis of variance (ANOVA) was carried out on
subjectsÕ raw scores using three factors. The factors were
the speaker (male and female), vowel (/a/ and /i/), and filter cut-off frequency (1.4, 2.0, 2.8, and 5.6 kHz). Table 3
shows the F and p values for each factor and for the various interactions. A three-way table of means is shown
in Table 4. For factors which varied within a session (4
filter conditions) a difference of greater than the least significant difference (LSD) of 4.26 percentage points within
a row of the table represents a statistically significant difference (at the 5% level). For factors which varied between
sessions (2 speakers · 2 vowels) a difference of greater
than the LSD of 7.16 percentage points within a column
of the table represents a statistically significant difference.
A. Simpson et al. / Hearing Research 210 (2005) 42–52
49
Fig. 2. Speech perception mean scores (filled circles) for subjects S1–S10 are indicated on the right y-axis as percentage correct. Error bars on each
symbol represent plus and minus one standard deviation from the mean. A speech banana for an average speech input level of 60 dBA is shown by
the shaded area. Aided thresholds (thin solid line) for each subject in dB SPL are shown on the left y-axis.
It is clear from this analysis that consonant scores
improved with increasing bandwidth. This increase
occurred with both speakers and with both vowel condi-
tions. The largest score increases were from 1.4 to 2.0,
and 2.8 to 5.6 kHz. There was no significant score increase from 2.0 to 2.8 kHz.
Table 3
Results of the analysis of variance of the consonant recognition results
Factor/s
F-value
p-value
Speaker
Vowel
Filter
Speaker * vowel
Speaker * filter
Vowel * filter
Vowel * filter * speaker
2.57
40.87
248.49
1.12
4.94
1.88
5.22
0.12
<0.001
<0.001
0.3
0.002
0.134
0.002
Note. Asterisk symbols refer to interactions among factors.
Table 4
SubjectsÕ mean speech perception results (% correct)
Speaker
Vowel
Low-pass filter cut-off frequency (kHz)
1.4
2.0
2.8
5.6
Female
/a/
/i/
42.81
34.22
65.78
49.38
66.09
46.56
72.97
53.59
Male
/a/
/i/
45.31
31.25
63.75
54.06
67.03
58.28
76.25
62.97
50
A. Simpson et al. / Hearing Research 210 (2005) 42–52
SubjectsÕ scores were significantly lower with the vowel /i/ for all filter conditions and for both speakers.
Scores were not significantly different between the two
speakers.
4. Discussion
Previous publications have reported that a small proportion of hearing-impaired listeners may show a decrease in speech understanding when the audibility of
high frequencies is increased (Hogan and Turner, 1998;
Murray and Byrne, 1986; Rankovic, 1991; Vickers
et al., 2001). However, in many of those studies, full
audibility of the speech signals could not be provided.
It has been difficult, therefore, to separate the effects of
limited signal audibility from the effects of cochlear dead
regions that may have been present in some of the subjects who participated in the studies. In the current study,
testing was conducted under conditions in which the
speech signals were highly audible to subjects. On average, these subjects showed a significant score increase
with increasing bandwidth. Nine out of the ten subjects
tested demonstrated a score increase with increasing
bandwidth. One subject showed no increases in speech
scores with increasing bandwidth beyond 2 kHz. None
of the subjects tested showed a score decrease with
increasing high-frequency audibility. The results seem
to indicate that whilst a worsening of performance is certainly possible when amplifying high frequencies, it
seems to occur in a very small number of cases, and
was not observed in the current study.
The question then remains of how to identify these
possible cases. One method that has been suggested is
the use of the TEN test, developed by Moore et al.
(2000a). Only three subjects (S01, S02, S06) in the current study showed dead regions according to the TEN
test. This result could partly be due to limitations of
the version of the TEN test used in the study. Even
the highest noise level was not sufficiently intense to
provide the required masking in the high frequencies.
In cases such as these, it was assumed that dead regions
existed where the listener had pure-tone thresholds of
90 dB HL or greater (Moore, 2001). Table 2 shows
the air conduction thresholds measured for all subjects.
In addition to S02 and S06, five subjects (S03, S04, S05,
S07, S09) had high frequency thresholds that were
greater than or equal to 90 dB HL. A threshold greater
than or equal to 90 dB HL was measured for S03 at
3000, 4000, and 8000 Hz; for S05 and S07, at 4000
and 8000 Hz; for S09 at 4000 Hz; and for S04 at
8000 Hz only. These subjects had possible high frequency dead regions based on the severity of their hearing thresholds.
One subject (S06) showed speech recognition results
in the consonant test that were consistent with the pres-
ence of dead regions found in the TEN test. Performance for this subject improved with increasing cut-off
frequency until the cut-off frequency was above the estimated edge frequency of the dead region. Further increases in the cut-off frequency resulted in no further
increases in speech recognition score. The remaining
nine subjects showed an increase in scores with increasing bandwidth.
Both Vickers et al. (2001) and Moore (2001) conclude
that there is little benefit in amplifying frequencies well
inside a dead region if that dead region extends from its
estimated edge frequency to the upper limit of hearing.
However, there may be some benefit in amplifying frequencies up to approximately one octave above the estimated edge frequency of a dead region. The highest
bandwidth used in the current study (5.6 kHz) was less
than one octave above the estimated edge frequency
of the assumed dead region for subjects S01 and S02
who both had dead regions according to the TEN test,
as well as for those subjects (S03, S04, S05, S07, S09)
with assumed dead regions at high frequencies based
on their hearing thresholds being greater than 90 dB
HL. In addition, the dead regions found for S01 and
S02 did not extend to the upper limit of these subjectsÕ
hearing.
It is therefore possible that subjectsÕ speech scores
may not have improved further if additional higher
cut-off frequencies had been tested. This would be difficult to evaluate, as it would not be possible to achieve
full audibility of the speech stimuli beyond 5.6 kHz given that stimuli were presented via a hearing aid. A
speech signal with very high gains would result in saturation of the hearing aidÕs output. It would be difficult to
determine whether a flattening of speech recognition
scores with further increases of bandwidth was due to
the presence of a dead region, limited audibility of the
speech signal, or some other cause (such as distortion
of the hearing aid).
To determine which phonetic features were affected
most by bandwidth, the responses of each subject were
analyzed to determine the percentage information
transmission for the phonetic features of voicing,
place, and manner (Miller and Nicely, 1955) as a function of filter cut-off frequency. The average results are
shown in Fig. 3. The percentage of transmitted information is plotted as a function of filter cut-off frequency for each feature. Information transmission
increases can be seen with increasing cut-off frequency
for both manner and place. Scores for both manner
and voicing remained high, even at the lowest cut-off
frequency of 1400 Hz. This indicates that information about these features can be extracted from lowfrequency components of speech. Place scores were
the lowest for the group at all frequencies. This finding
is consistent with those of other studies (Dubno et al.,
1982).
A. Simpson et al. / Hearing Research 210 (2005) 42–52
51
frequencies may be impractical for everyday use of a
hearing aid, and could result in discomfort for the listener, and/or acoustic feedback. Alternative signal processing, such as frequency compression, may be able to
provide these listeners with additional high-frequency
information while avoiding these problems (Simpson
et al., 2005).
Acknowledgements
Fig. 3. Average information transmission scores for all subjects. The
features of manner (open circles), place (filled triangles), and voicing
(filled circles) are plotted as a function of the low-pass filter cut-off
frequency. Error bars represent plus and minus one standard deviation
from the mean.
5. Conclusions
In the current study, testing was conducted under
conditions in which the speech signals were highly audible to subjects within four different bandwidths delimited by low-pass filters with varying cut-off frequencies.
The TEN test was carried out with all subjects. For
one subject (S06), extensive dead regions were found
with the TEN test. For that subject, speech recognition
scores did not improve with increasing bandwidth. It is
likely that many of the remaining subjects also had dead
regions above 3 kHz based on their hearing thresholds,
but these could not be demonstrated by means of the
TEN test. On average, these subjects showed a significant score increase with increasing bandwidth. None
of the subjects tested showed a score decrease with
increasing high frequency audibility.
Before considering reducing high-frequency amplification when fitting a hearing aid, the clinician should
consider whether the listener demonstrated extensive
dead regions that extended to the upper limit of hearing.
In addition, when a dead region is found with the TEN
test, it is important that amplification is provided to at
least one octave above the edge frequency of the dead
region (Baer et al., 2002; Vickers et al., 2001).
Certainly many listeners with severe high-frequency
losses are able to make some use of high-frequency
speech cues if these cues can be made audible. The
average high-frequency gain at 1000, 2000, and
4000 Hz was calculated for the present study to equal
54 dB (2cc coupler) for the subject group. In comparison, the NAL-NL1 (Byrne et al., 2001) fitting guideline
prescribes an average of 40 dB gain (2cc coupler) at the
same high frequencies for this group of subjects. The
levels of audibility achieved in the experiments at these
The authors are grateful for the financial support of
the Commonwealth of Australia through the Cooperative Research Centres program. The second author received financial support from the Garnett Passe and
Rodney Williams Memorial Foundation. Aided thresholds were measured using Presentation software
version 0.70 which can be downloaded from www.neurobs.com. MACarena Speech Tests were developed by
Dr. WaiKong Lai from the University of Zurich. The
comments of an anonymous reviewer and Prof. Brian
Moore on a previous version of this manuscript are
greatly appreciated. Thanks are due also to the Statistical Consulting Centre at the University of Melbourne,
and Adam Hersbach. We are especially grateful to the
hearing-aid users who participated as subjects in the
experiments.
References
Baer, T., Moore, B.C., Kluk, K., 2002. Effects of low pass filtering on
the intelligibility of speech in noise for people with and without
dead regions at high frequencies. J. Acoust. Soc. Am. 112, 1133–
1144.
Bentler, R.A., Pavlovic, C.V., 1989. Transfer functions and correction
factors used in hearing aid evaluation and research. Ear Hear. 10,
58–63.
Byrne, D., Dillon, H., Tran, K., Arlinger, S., Wilbraham, K., Cox,
R.M., Hagerman, B., Hetu, R., Kei, J., Lui, C., Kiessling, J.,
Nasser Kotby, M., Nasser, H.A., El Kholy, W., Nakanishi, Y.,
Oyer, H., Powell, R., Stephens, D., Meredith, R., Sirimanna, T.,
Tavartkiladze, G., Frolenkov, G., Westerman, S., Ludvigsen, C.,
1994. An international comparison of long-term average speech
spectra. J. Acoust. Soc. Am. 96, 2108–2120.
Byrne, D., Dillon, H., Ching, T., Katsch, R., Keidser, G., 2001. NALNL1 procedure for fitting nonlinear hearing aids: characteristics
and comparisons with other procedures. J. Am. Acad. Audiol. 12,
37–51.
Carhart, R., Jerger, J.F., 1959. Preferred method for clinical determination of pure-tone thresholds. J. Speech Hear. Disord. 24, 330–
345.
Dubno, J.R., Dirks, D.D., Langhofer, L.R., 1982. Evaluation of
hearing-impaired listeners using a nonsense-syllable test. II. Syllable recognition and consonant confusion patterns. J. Speech Hear.
Res. 25, 141–148.
Dybala, P.D., Thibodeau, L., 2002. Examination of the relevance of
cochlear dead regions to hearing aid fittings. J. Acoust. Soc. Am.
112, 2356.
Fletcher, H., 1953. Speech and Hearing in Communication. Van
Nostrand, New York.
52
A. Simpson et al. / Hearing Research 210 (2005) 42–52
Fletcher, H., Galt, R.H., 1950. The perception of speech and its
relation to telephony. J. Acoust. Soc. Am. 22, 89–151.
Hogan, C.A., Turner, C.W., 1998. High-frequency audibility: benefits for hearing-impaired listeners. J. Acoust. Soc. Am. 104, 432–
441.
Markham, D., Hazan, V., 2002. The UCL speaker database. Speech,
Hearing and Language: UCL Work in Progress, 14, 1–17.
Miller, G., Nicely, P., 1955. An analysis of perceptual confusions
among some English consonants. J. Acoust. Soc. Am. 27, 623–637.
Moore, B.C., 2001. Dead regions in the cochlea: diagnosis, perceptual
consequences, and implications for the fitting of hearing aids.
Trends Amplif. 5, 1–34.
Moore, B.C., 2002. Response to ‘‘Articulation index predictions for
hearing-impaired listeners with and without cochlear dead
regions’’. J. Acoust. Soc. Am. 111, 2549–2550.
Moore, B.C.J., Glasberg, B.R., 1997. A model of loudness perception
applied to cochlear hearing loss. Aud. Neurosci. 3, 289–311.
Moore, B.C., Glasberg, B.R., 1998. Use of a loudness model for
hearing-aid fitting. I. Linear hearing aids. Br. J. Audiol. 32, 317–
335.
Moore, B.C., Huss, M., Vickers, D.A., Glasberg, B.R., Alcantara, J.I.,
2000a. A test for the diagnosis of dead regions in the cochlea. Br. J.
Audiol. 34, 205–224.
Moore, B.C.J., Huss, M., Glasberg, B.R., 2000b. Diagnosis of Dead
Regions. Starkey Laboratories, Inc., Eden Prairie, MN.
Murray, N., Byrne, D., 1986. Performance of hearing-impaired and
normal hearing listeners with various high frequency cut-offs in
hearing aids. Aust. J. Audiol. 8, 21–28.
National Acoustic Laboratories, 2000. Speech and Noise for Hearing
Aid Evaluation. Australian Hearing, Chatswood, NSW.
Rankovic, C.M., 1991. An application of the articulation index to
hearing aid fitting. J. Speech Hear. Res. 34, 391–402.
Rankovic, C.M., 2002. Articulation index predictions for hearingimpaired listeners with and without cochlear dead regions. J.
Acoust. Soc. Am. 111, 2545–2548, author reply 2549–50.
Simpson, A., Hersbach, A.A., McDermott, H.J., 2005. Improvements
in speech perception with an experimental nonlinear frequencycompression hearing device. Int. J. Audiol. 44, 281–292.
Steinberg, J.C., Gardner, M.B., 1937. The dependence of hearing
impairment on sound intensity. J. Acoust. Soc. Am. 9, 11–23.
Stelmachowicz, P.G., Pittman, A.L., Hoover, B.M., Lewis, D.E., 2001.
Effect of stimulus bandwidth on the perception of /s/ in normaland hearing-impaired children and adults. J. Acoust. Soc. Am. 110,
2183–2190.
Thornton, A.R., Abbas, P.J., 1980. Low-frequency hearing loss:
perception of filtered speech, psychophysical tuning curves, and
masking. J. Acoust. Soc. Am. 67, 638–643.
Vickers, D.A., Baer, T., Moore, B.C.J., 2001. Effects of lowpass
filtering on speech intelligibility for listeners with dead regions at
high frequenies. Br. J. Audiol. 35, 148–149.
Original Article
International Journal of Audiology 2006; 45:619 629
Andrea Simpson*,%
Adam A. Hersbach*,$
Hugh J. McDermott*,%
*The Cooperative Research Centre for
Cochlear Implant and Hearing Aid
Innovation, Melbourne, Australia
$
The Bionic Ear Institute, Melbourne,
Australia
%
Department of Otolaryngology, The
University of Melbourne, Australia
Key Words
Steeply sloping hearing loss
Hearing impairment
Hearing aid users
Frequency compression
Frequency-compression outcomes in listeners
with steeply sloping audiograms
Resultados de la compresión de frecuencia en oyentes
con audiogramas de pendiente pronunciada
Abstract
Sumario
Previous investigation of an experimental, wearable
frequency-compression hearing aid revealed improvements in speech perception for a group of listeners with
moderately sloping audiograms (Simpson et al, 2005). In
the frequency-compression hearing aid, high frequencies
(above 1600 Hz) were amplified in addition to being
lowered in frequency. Lower frequencies were amplified
without frequency shifting. In the present study, an
identical frequency-compression scheme was evaluated
in a group of seven subjects, all of whom had steeply
sloping hearing losses. No significant differences in group
mean scores were found between the frequency-compression device and a conventional hearing instrument for
understanding speech in quiet. Testing in noise showed
improvements for the frequency-compression scheme for
only one of the five subjects tested. Subjectively, all but
one of the subjects preferred the sound quality of the
conventional hearing instruments. In conclusion, the
experimental frequency-compression scheme provided
only limited benefit to these listeners with steeply sloping
hearing losses.
Las investigaciones previas sobre un auxiliar auditivo
experimental con compresión de frecuencia revelaron
una mejor percepción del lenguaje en un grupo de oyentes
con audiogramas de pendiente moderada (Simpson y
col., 2005). En el auxiliar auditivo con compresión de
frecuencia, las frecuencias agudas (por encima de
1600Hz) se amplificaron además de ser reducidas
en frecuencia. Las frecuencias más graves fueron
amplificadas sin cambio en la frecuencia. En el presente
estudio, un esquema similar de compresión de frecuencia
fue evaluado en un grupo de 7 sujetos, todos los
cuales tenı́an hipoacusia con pendientes abruptas. No se
encontraron diferencias significativas en los puntajes
medios de grupo para entender el lenguaje en silencio
entre el dispositivo con compresión de frecuencia y el
instrumento convencional. La evaluación en ruido
mostró mejorı́a para el esquema de compresión de la
frecuencia en sólo uno de los cinco sujetos evaluados.
Subjetivamente, todos menos uno de los sujetos,
prefirieron la calidad del sonido de los instrumentos
auditivos convencionales. En conclusión, el esquema
experimental de compresión de frecuencia aportó sólo
un beneficio limitado para los oyentes con hipoacusias de
pendiente abrupta.
Hearing impairment results in a decrease in audibility due to
threshold elevation. The greater the hearing loss, the larger the
amount of amplification needed to provide the listener with
adequate audibility of speech sounds. This may not be possible
in individuals with a severe loss in the high frequencies, but who
have good hearing sensitivity in the low frequencies. Conventional hearing aids are limited in their ability to provide
high-frequency amplification without feedback occurring. In
addition, these high levels of amplification may be undesirable as
they could result in either distortion of the speech signal or
discomfort for the listener. There is some evidence to suggest
that a small number of listeners with severe high-frequency loss
show no benefit from high-frequency amplification (Hogan &
Turner, 1998; Murray & Byrne, 1986; Rankovic, 1991). What
remains unclear is why only certain individuals show this lack of
benefit. One explanation is that of ‘dead regions’. The term dead
region was coined to define a part of the cochlea with nonfunctioning inner hair cells (Moore & Glasberg, 1997; Thornton
& Abbas, 1980). Previous studies have suggested that individuals
with no dead regions often obtain some benefits from highfrequency amplification, whereas individuals with dead regions
may receive little additional information if the amplified signals
are well within an extensive dead region (Baer et al, 2002;
Vickers et al, 2001). Dead regions can be diagnosed from
psychophysical tuning curves or with the TEN test (Moore et
al, 2000; Moore, 2001). Moore (2001) also states that dead
regions are extremely likely once hearing thresholds exceed 90
dB HL. All these factors result in many individuals with severe
losses not obtaining adequate performance from conventional
hearing aids. A possible solution in these cases would be to
provide additional high-frequency information by making use of
the listener’s low-frequency hearing (Moore, 2001).
Various sound-processing schemes have been designed with
the aim of presenting information from high frequency components of speech at lower frequencies. These schemes often used
non-linear modulation techniques to shift high-frequency speech
components downwards to a lower frequency range. The downward shift was typically disproportionate, meaning that frequency ratios contained in the spectral information were not
preserved during processing. The resulting signals were mixed
with those obtained from lower frequencies. One scheme
(Johansson, 1961) implemented disproportionate frequency
ISSN 1499-2027 print/ISSN 1708-8186 online
DOI: 10.1080/14992020600825508
# 2006 British Society of Audiology, International
Society of Audiology, and Nordic Audiological Society
Andrea Simpson
The Cooperative Research Centre for Cochlear Implant and Hearing Aid
Innovation, 384-388 Albert Street, East Melbourne 3002, Australia
Email: [email protected]
Accepted:
May 8, 2006
shifting in which high-frequency energy was passed through a
non-linear modulator that converted it into low-frequency noise.
Several studies of these early types of frequency-lowering
methods (Ling, 1968; Johansson, 1961; Velmans, 1974; Velmans
& Marcuson, 1983) reported little success in providing speech
understanding benefits. This may have happened for a variety of
reasons. Much of the information regarding the spectral shape of
the incoming signal was lost as a result of the processing
technique. As mentioned, those methods did not preserve
frequency ratios in the high frequencies when these were shifted
to lower frequencies. In addition, those schemes may have
provided some additional high-frequency information at the
expense of other perceptual cues by overlapping the shifted and
un-shifted signals.
Proportional frequency shifting offers the advantage that the
ratios among frequency components are not changed by the
processing. This appears to be particularly important for the
recognition of vowels (Neary, 1989). It has been shown that
normally-hearing listeners can understand frequency-lowered
speech provided that the amount of proportional frequency
shifting is not too large. Generally, studies have shown that the
performance of normally- hearing listeners remained high when
signals were compressed in frequency such that the output
bandwidth was no narrower than 70% of the original bandwidth
(Daniloff et al, 1968; Fu & Shannon, 1998).
This type of scheme was tested by Turner and Hurtig (1999).
Fifteen listeners with high-frequency hearing impairment identified nonsense syllables that were uniformly lowered in frequency
by factors ranging from 0.5 to 1. The scheme produced
significant improvements in intelligibility with female-talker
speech material for about half the listeners when the frequency
compression factor was set to 0.8.
Proportional frequency shifting has also been attempted with
hearing-impaired listeners using a ‘slow play’ method (Bennett &
Byers, 1967; McDermott et al, 1999; McDermott & Knight,
2001). Segments of the speech signal are recorded and then
played back at a lower speed than employed for recording. Two
commercially available hearing instruments that incorporate this
type of frequency-lowering processing are the TranSonic and
ImpaCt DSR675 (AVR Communications Ltd.). In these devices,
the activation of frequency shifting is dependent upon
the incoming signal. Incoming signals dominated by components at frequencies above 2500 Hz are shifted down by a
factor that is programmable for each listener. If the input signal
is not dominated by frequencies above 2500 Hz, then all signals
are amplified with no frequency shifting. Parent et al (1998)
found that two of the four adult subjects tested demonstrated
speech perception benefits with the TranSonic device relative to
their conventional devices. Positive outcomes have also
been reported when the TranSonic was fitted to a small
number of hearing-impaired children (Davis-Penn & Ross,
1993; Rosenhouse, 1990).
However, the benefits provided by the TranSonic could be the
result of providing additional low-frequency amplification,
rather than frequency-lowering per se. McDermott et al (1999)
reported higher scores for five adult subjects using the TranSonic
than for their own hearing aids. However, only two of those
subjects appeared to obtain additional speech information
specifically from the high-frequency signal components after
they were lowered. An additional study (McDermott & Knight,
2001) of the AVR ImpaCt hearing instrument reported little
difference in performance between the ImpaCt aid and the
subjects’ own conventional aids.
The perceptual performance of a number of linear and
nonlinear frequency-compression schemes was evaluated by
Reed et al (1983). In a preliminary study, six subjects with
normal hearing participated in experiments that investigated
whether any of the schemes could improve the discriminability of
consonant stimuli. Although none of the schemes provided
better performance than a standard condition that applied only
low-pass filtering to the stimuli, the best scheme was found to be
a variant that progressively increased the amount of frequency
compression for input frequencies above approximately 1200 Hz.
Lower input frequencies were hardly changed by the processing.
Subsequently, this scheme was one of two variants tested with a
small group of hearing-impaired subjects (Reed et al, 1985).
None of the three subjects who completed a consonant
identification test with the frequency-compression scheme obtained higher scores than with linear amplification. Overall,
these findings suggested that, although frequency compression
did not provide a perceptual benefit, the scheme that resulted in
the best scores for the consonant tests applied increasing
amounts of frequency lowering to relatively high input frequencies, while leaving lower frequencies unchanged.
Recently, the performance of an experimental frequencycompression hearing device was investigated by Simpson et al
(2005). The input to output frequency relationship of the scheme
is represented in Figure 1. The scheme divided incoming signals
into two broad bands based on a chosen cut-off frequency, which
ranged from 1600 2500 Hz. Signal components below the cut-off
frequency were amplified with appropriate frequency shaping
and amplitude compression but without frequency shifting.
Signal components above the cut-off were compressed in
frequency in addition to undergoing amplification. The frequency compression was non-linear and applied progressively
larger shifts to components having increasingly high frequencies.
Consequently, a wide range of high-frequency input signals
resulted in a narrower range of output signals. A possible
advantage of the scheme is that there is no spectral overlap
between the shifted and un-shifted signals. This results in the
first-formant and most of the second-formant frequency range
remaining preserved, as low- and mid-frequency information are
not shifted with the device. A possible disadvantage of the
experimental scheme is that it does not preserve frequency ratios
for those high frequencies that are compressed. It is a possibility
that the perception of certain sounds, such as music, may be
affected adversely. In addition, the fitting parameters of the
frequency-compressed signal, such as cut-off frequency and gain,
are based on the auditory characteristics of the subjects’ better
ear. This signal is then presented to both ears. The experimental
scheme may thus be suboptimal for cases where subjects have
asymmetrical hearing thresholds at high frequencies.
Seventeen hearing-impaired subjects with moderate-to-severe
sensorineural hearing loss and sloping audiograms wore the
device in place of their conventional hearing instruments for
approximately five weeks. Their recognition of monosyllabic
words was evaluated with each instrument. Eight of the 17
subjects obtained a significant score improvement (p B/0.05)
when using frequency compression. The remaining subjects had
similar scores with each device, with the exception of one subject
620
International Journal of Audiology, Volume 45 Number 11
Figure 1. The input-to-output frequency mapping function for the frequency compression scheme for two cut-off frequencies of
1250 Hz and 1600 Hz.
whose score decreased significantly with frequency compression.
In summary, the experimental frequency-compression scheme
proved successful in providing additional high-frequency
information for some listeners with moderately sloping losses,
whose hearing thresholds became severe to profound in the
2000 8000 Hz region.
In the study of Simpson et al (2005), the experimental
frequency-compression scheme was not tested with listeners
with steeply sloping hearing losses. Typically, these individuals
have near-normal hearing thresholds in the low frequencies and
a profound hearing loss for frequencies above approximately
1500 Hz. The effects of frequency transposition with this type of
hearing loss have been investigated previously with a group of six
subjects (McDermott & Dean, 2000). Proportional frequency
shifting was applied to monosyllabic words in which all
frequencies were lowered by a factor of 0.6. Subjects underwent
extensive auditory training. Disappointingly, frequency compression had no significant effect on speech identification.
The current study examined whether the experimental frequency-compression scheme described by Simpson et al (2005)
would benefit listeners with steeply sloping hearing losses. All
subjects were initially fitted with a conventional device. Speech
perception comparisons in both quiet and noise were made
between this device and the frequency-compression device. The
APHAB questionnaire (Cox & Alexander, 1995) was given to
subjects at the end of the trial in order to determine their
subjective preferences.
Method
Initially, an informal listening task was carried out to determine
whether subjects with steeply sloping losses would prefer certain
frequency-compression parameters. There are two main parameters that can be adjusted in the device. The first is the point at
which frequency-compression begins, or the cut-off frequency.
Frequency-compression outcomes in
listeners with steeply sloping audiograms
The second parameter is the degree of frequency compression, or
the slope of the input frequency to output frequency relationship. Both of these parameters were adjusted in a variety of ways
in an attempt to determine an acceptable fitting for this group of
subjects. The parameters selected for further investigation were
cut-off frequencies of 1000, 1250, and 1600 Hz, and compression
slopes of 4:1, 2:1, and 0.5:1.
Ten hearing-impaired adults with steeply sloping losses
participated in a pilot study. A paired comparison method was
used to determine which settings subjects preferred. Of the 10
subjects, only four consistently chose the same parameter
combination. In general, it appeared that subjects preferred
the higher cut-off frequencies, combined with an input to output
frequency compression slope of 2:1. These preferred parameters
(cut-off frequency of 1250 or 1600 Hz, and a slope of 2:1) were
then fixed when fitting the frequency-compression hearing aid,
as detailed below.
Subjects
Seven hearing-impaired adults, three women and four men,
participated in the study. Relevant information about them is
shown in Table 1. Four of the subjects were experienced hearingaid users (S32, S37, S38, S39). These subjects had worn hearing
aids for several years and wore their current hearing aids on a
daily basis. The remaining three subjects had not worn hearing
aids previously (S34, S35, S36). Their hearing threshold levels,
measured conventionally under headphones, are shown in
Figure 2. For all subjects, hearing losses were assumed to have
primarily a sensorineural origin, based on standard air- and
bone-conduction audiometry. Subjects were not tested for
cochlear dead regions. However, as discussed later, the audiogram configurations suggest that dead regions were present at
high frequencies in most, if not all ears. Subjects were not paid
for their participation in the experiments, although expenses
such as travel costs were reimbursed. The participation of the
Simpson/Hersbach/McDermott
621
Table 1. Relevant information about the subjects who participated in the study, and their hearing aids
Subject
Age (yrs) Sex
Probable Etiology
of Hearing Loss
Type of Own
Hearing Aid/s
S32
75
M
Unknown
Widex Senso Diva
S34
S35
S36
S37
S38
33
33
57
53
68
F
M
M
M
F
Unknown
Noise exposure
Hereditary
Hereditary
Unknown
Non hearing aid user
Non hearing aid user
Non hearing aid user
Miracle Ear
Siemens
S39
72
F
Ototoxic drugs
Bernafon RB15
Features of Own Aid(s)
Processing Strategy
of Own Aid(s)
Digital, adaptive directional
microphone
Analog, omnidirectional
Digital, adaptive directional
microphone
Digitally programmable,
omnidirectional
Compression
Linear
Compression
Linear
Each subject’s medical history and hearing-aid usage were
documented during the first test session. A pure-tone audiogram, including both air- and bone conduction, was obtained,
and where appropriate the electro-acoustic characteristics of
each subject’s own hearing aids were measured and recorded.
Table 1 shows relevant details of each subject’s own hearing aids
and hearing-aid experience.
Each subject was fitted with identical conventional hearing
instruments (Phonak Supero 412) using the manufacturer’s
fitting software. The Supero 412 is a behind-the-ear (BTE) digital
power instrument. It is specified to have a maximum output and a
maximum gain of approximately 140 dB SPL and 80 dB, respec-
tively (measured in an ear simulator). The gain and amplitude
compression characteristics are separately adjustable in five
partially-overlapping frequency bands. The amplitude compression attack and release times are 5 and 30 ms respectively.
User-selectable normal and noise-reduction programs were
created by entering the subjects’ pure-tone thresholds into the
fitting software. The initial fitting suggestion was based on the
NAL-NL1 fitting guideline (Byrne et al, 2001). When necessary,
these settings were altered at the follow-up sessions based on
subject feedback. For example, the three most common subject
reports regarded their own voice, loudness discomfort, and
general sound quality. Adjustments were made to the lowfrequency gain of the device in steps of 3 dB if the subject was
dissatisfied with the sound of their own voice. If the subject
reported discomfort when listening to loud noises, the overall
maximum power output of the hearing aid was reduced in steps
of 3 dB. Typically, changes of no greater than two steps were
applied when these adjustments were required. If the subject
reported dissatisfaction with the general sound quality of the
device and they were a previous hearing-aid user, the programming of the device was adjusted to approximate the amplification characteristics of that subject’s own hearing aids, based on
2-cm3 coupler measurements.
For those subjects who had worn hearing aids previously,
speech perception comparisons were made between the conventional device and the subjects’ own hearing aids. Testing was
carried out using open-set CNC monosyllabic word lists, presented from a loudspeaker at an intensity level of 65 dBA. Subjects
were included in the current study only if their speech perception
scores with the conventional hearing device were the same as, or
better than, those obtained with their own hearing aids. Each
subject had been wearing the conventional hearing devices for 4 5
weeks prior to the commencement of the current study.
The frequency-compression device was fitted in a manner
similar to that described in Simpson et al (2005). Initially, each
subject’s fitting parameters, determined as described above, were
programmed into the modified conventional hearing devices
attached to the SHARP processor. For each subject, the loudness
of the test signals specified above was approximately equalized
across frequency. To achieve this, amplification of the high
frequencies was adjusted by means of a loudness rating method.
The subject was seated, wearing the frequency-compression
device, at a distance of 1 m directly in front of the loudspeaker.
Subjects were given a categorical list of nine loudness levels.
622
International Journal of Audiology, Volume 45 Number 11
subjects in these experiments was in accordance with the
‘Guiding principles for research involving human or animal
subjects’, and was approved by the Human Research Ethics
Committee of the Royal Victorian Eye and Ear Hospital,
Melbourne, Australia.
Procedure
The hardware of the frequency-compression hearing aid has
been described previously (Simpson et al, 2005). It consisted of
two main parts: a pair of modified behind-the-ear (BTE)
conventional hearing devices, and a SHARP programmable
body-worn speech processor (Zakis et al, 2004). Sound entered
the system via the microphone of each BTE, and was processed
by the conventional hearing device. The BTEs were modified so
that their outputs were directed to the SHARP processor. The
SHARP processor then manipulated the signals to perform
frequency compression and sent the outputs to the earphone
receivers located in the BTEs.
Device fitting stimuli
The frequency-compression device was fitted by making use of
equal-loudness level measurements. Third-octave narrowband
noises with duration 0.5 s, were selected. Each noise stimulus was
separated from the next stimulus by a silent interval of duration
0.5 s. The noises were ramped in level at each end with linear
ramps of duration 30 ms. The level of each noise was set to equal
the average 1/3-octave level at the corresponding frequency of
the international long-term average speech spectrum (Byrne et
al, 1994) for an overall level of 70 dB SPL measured at the
subject’s listening position.
Device fitting method
Figure 2. Hearing threshold levels (dB HL) for the subjects who participated in the study. Left ear thresholds for each subject are
shown by the solid line. Thresholds obtained for the right ear are shown by the dotted line. The hatched area represents frequencies
which were compressed. The lower edge of the hatched area represents the final cut-off frequency for each subject. Thresholds of
subjects who were fitted with a cut-off frequency of 1600 Hz are shown on the left of the figure, and those of subjects who were fitted
with a cut-off frequency of 1250 Hz are shown on the right.
The categories were: very soft, soft, comfortable but slightly
soft, comfortable, comfortable but slightly loud, loud but OK,
uncomfortably loud, extremely uncomfortable, and painfully
loud. The subject was asked to indicate the loudness of the noise
stimuli by pointing to a category on the printed list. Frequency
compression was enabled in the device before presenting narrowband noises at 1000, 1250, 1600, 2000, 2500, 3100, and 4000 Hz.
Subjects were thus making a loudness judgment based on the
compressed signal. The amplification of each frequency was
adjusted manually by the clinician for each narrow-band noise
Frequency-compression outcomes in
listeners with steeply sloping audiograms
Simpson/Hersbach/McDermott
623
until the subject reported each noise to be ‘comfortable but
slightly soft’. Due to the severity of hearing loss for most
subjects at these frequencies, it was often not possible to achieve
this level at all frequencies without feedback oscillation occurring. In these cases, the level was set to the maximum possible
without feedback oscillation.
Initially a cut-off frequency of 1250 Hz was chosen for each
subject. After fitting, subjects were asked to wear the device away
from the laboratory and provide feedback about it at two followup sessions. If the subject was dissatisfied with the sound quality
of the device, the output level and the cut-off frequency were
adjusted at these sessions. The clinician first reduced the level of
the frequency-compressed signal by 3 5 dB. If the subject
continued to report an unacceptable sound quality, then the cutoff frequency was adjusted to 1600 Hz. Figure 2 shows the final
cut-off frequencies selected for each subject. Three subjects (S32,
S34, S39) were fitted with a cut-off frequency of 1250 Hz, and four
subjects (S35, S36, S37, S38) were fitted with a cut-off frequency of
1600 Hz. As mentioned above, the same frequency-compression
slope was selected for all subjects. No further adjustments were
made to the program for the remainder of the trial.
Speech test stimuli and methods
The speech tests comprised recognition of open-set monosyllabic
words, closed-set medial consonants in quiet, and open-set
sentences presented with a competing noise. For all evaluations
of speech intelligibility, each subject was tested individually in a
medium-sized sound-attenuating booth. The volume control on
each subject’s conventional hearing device was set so that speech
at normal conversational levels in quiet would be comfortably
loud. For most subjects, this was the default volume control
setting. This setting was noted and fixed for all following test
sessions.
Each subject was given the frequency-compression device for
a total period of 4 6 weeks. All testing was carried out during
the final two sessions of this period. Two sets of scores for each
of the two conditions (i.e. with frequency compression, and with
conventional aids) for each test were averaged and compared.
Word recognition in quiet
STIMULI
The stimuli consisted of consonant- vowel nucleus -consonant
(CNC) monosyllabic words (Peterson and Lehiste, 1962). They
were presented from audio recordings of a male speaker with a
typical Australian accent. There were a total of 50 words per list.
The order in which lists were presented to subjects across
sessions was randomized. No lists of words (other than practice
lists) were repeated for any subject during the trial. When
measured at the subject’s listening position (about 1 m from the
loudspeaker), the average level of the words was maintained at
55 60 dBA across subjects and across test sessions. These levels,
which are similar to the levels of speech in normal conversation,
were generally perceived as comfortably loud when heard by the
subjects through their hearing aids.
subjects with the testing procedure and materials. Subjects
were instructed to repeat each word immediately after hearing
it, and to guess if unsure. Responses from the practice list
were excluded from the data analysis. After the practice list, a
further two lists were presented for each of the two test
conditions. Subjects’ responses were analysed to determine the
number of phonemes correctly recognized out of a total of 150
phonemes per list.
Consonant recognition in quiet
STIMULI
Vowel-consonant-vowel (VCV) utterances were used for the test
of consonant recognition. They were recorded by a male
Australian speaker. The 16 stimuli were: /p/, /t/, /k/, /b/, /d/, /g/
, /m/, /n/, /s/, /sh/, /z/, /f/, /v/, /ch/, /j/, and /th/. The surrounding
vowels were /a/ in both the initial and final positions. The levels
of the tokens were approximately equalized. The average levels
were approximately 60 dBA at the subject’s listening position, a
distance of 1 m from the loudspeaker. Each stimulus was
repeated six times making a total of 96 token presentations.
METHOD
A closed-set procedure was used for the test of consonant
recognition, in which 16 buttons appeared on a computer screen.
The buttons were of the following format: vCv, where the vowel
was /a/, and C represented each consonant.
The test consisted of three blocks in which each token was
heard twice by the subject. The stimuli were presented in a
random order in each subtest. The subjects were presented with
the 16 tokens in the test set, as displayed on the computer screen.
They were instructed to identify which consonant they heard
after each stimulus was presented, by pressing the corresponding
button on the screen. There was no option for a response other
than one of the 16 consonants in the test set. At the start of each
test session, each subject carried out one block to gain practice
and confidence in the procedure. The results were recorded as
usual, but disregarded in the final analysis of the data.
Sentence recognition in noise
STIMULI
The sentence test material consisted of CUNY sentence lists
(Boothroyd et al, 1985). There were a total of 60 lists available.
Each list consisted of 12 sentences (i.e. approximately 102
words). The material, which was recorded by a female speaker
with a typical Australian accent, was presented through a
loudspeaker at an average level of approximately 65 dBA. As
described below, an adaptive procedure was used to estimate the
signal-to-noise ratio (SNR) for a target score of 50% words
correct. For this procedure, the competing noise was eight-talker
babble. This noise was mixed at a controlled level with the
sentence material using a two-channel audiometer. During the
testing, the sentence lists were selected at random, and no list
was repeated for any subject throughout the test sessions.
METHOD
METHOD
For each of two conditions: (1) using the conventional hearing
device; and (2) using the frequency-compression scheme, a
practice CNC words list was presented first to familiarize
Initially one practice list was presented, and the subject was
instructed to repeat as many words as they could identify after
hearing each sentence, and to guess if unsure. A further four lists
624
International Journal of Audiology, Volume 45 Number 11
were then presented, and one point was given for each correct
word repeated by the subject. The total was calculated as
a percentage score. This score was treated as the individual
subject’s highest potential score in quiet. This score had to be
equal to or greater than 75% for testing to continue.
Subjects with scores of less than 75% in quiet were not tested
further.
Subsequently, sentences were presented combined with noise,
starting at a signal-to-noise ratio (SNR) of 10 dB. Although
pauses were inserted between sentences to enable scores to be
calculated, the noise was presented continuously throughout the
testing procedure. After the subject’s response to each pair of
consecutive sentences (containing approximately 9 13 words),
the proportion of words correctly identified was determined. The
level of the noise was either decreased or increased, according to
whether the subject’s score was respectively less or greater than
50% correct. Initially, changes in the noise level were made in
steps of 5 dB, and each change in the direction of noise-level
adjustments was noted. After two such reversals were obtained,
subsequent changes in the noise level were made in steps of 3 dB.
The test proceeded until a further 10 reversals were obtained.
The final SNR was the average across the last 10 reversals.
Subjective assessment
The Abbreviated Profile of Hearing Aid Benefit (Cox and
Alexander, 1995) was used to measure subjective benefit for
each device type. The APHAB is a 24-item self-assessment scored
on four subscales (i.e. with six items per subscale). Three
subscales, Ease of Communication, Reverberation, and Background Noise address speech understanding in everyday life. The
fourth subscale, Aversiveness to Sounds, assesses negative reac-
tions to environmental sounds. At the second-last testing session,
subjects were given the APHAB to take home to complete.
Results
Word and consonant recognition in quiet
For the CNC words test, the mean phoneme scores obtained by
each subject with their conventional hearing devices and with the
frequency-compression device are shown in Figure 3. Across the
seven subjects, these scores were compared between the two
hearing-aid processing schemes by means of a two-factor
analysis of variance (ANOVA). No statistically significant
difference was found between the frequency-compression scheme
(FrC) and the conventional hearing device (CD) (p /0.11). In
addition, there was no significant interaction between the
scheme and the subject factors (p /0.8).
Figure 3 also shows the mean percentage correct scores
obtained by each subject for the CD and the FrC device for
the consonant test. A two-factor ANOVA was carried out on the
subjects’ scores. Across all subjects there was no statistically
significant difference between scores for the conventional hearing device and the frequency-compression device (p/0.186).
However, there was a significant interaction term (subject /
scheme, p/0.002). Therefore, each subject’s data were analysed
separately with pair-wise comparisons using the Holm-Sidak
method (Hochberg and Tamhane, 1987). Although visual
inspection shows higher individual scores with the frequencycompression device than with the conventional device for S35,
S36, and S37, this difference was significant only for S37 (t/
2.736, p/0.011). Subjects S32, S34, and S38 show individual
lower scores with the frequency-compression device than with
Figure 3. Mean percent correct scores for open-set recognition of phonemes (in CNC monosyllabic words) and for closed-set
identification of consonants (in vCv format) obtained by the seven hearing-impaired subjects who participated in the study. Unfilled
columns show phoneme scores obtained using the conventional hearing devices (CD: Phonemes), black filled columns show phoneme
scores obtained using the experimental frequency-compression scheme (FrC: Phonemes), hatched columns show consonant scores
obtained with the conventional device (CD: Consonants), and grey filled columns show consonant scores obtained with the
frequency-compression scheme (FrC: Consonants). Scores averaged across subjects are shown in the group of right-most columns,
with error bars indicating one standard deviation. Statistical significance is shown by asterisk symbols: * 0.01 B/pB/0.05, ** 0.001 B/
pB/0.01, *** pB/0.001.
Frequency-compression outcomes in
listeners with steeply sloping audiograms
Simpson/Hersbach/McDermott
625
the conventional device, and this was significant for subjects S32
(t/3.719, pB/0.001) and S34 (t/2.177, p/0.038).
Sentence recognition in noise
The SNRs obtained by each subject from the adaptive test of
sentence comprehension in noise are shown in Figure 4. Note
that larger SNR values result from poorer performance on this
test; therefore, shorter bars in the graph indicate better
performance. As indicated on the vertical axis, five subjects
participated in these tests. The excluded subjects were S32 and
S39, who scored less than 75% words correct in the initial
sentence test without competing noise. Although some variability in the results from this test was found among the subjects,
a two-factor ANOVA revealed that the effect of aid type was
statistically significant (df/1, t/2.532, p/0.02). A significant
interaction term (subject /scheme, pB/0.001) was also found.
An analysis of each subject’s data separately with pair-wise
comparisons (Holm-Sidak method) showed that only one
subject (S35) improved significantly (t/5.233, pB/0.001) when
using frequency-compression. To further investigate, the ANOVA was repeated with S35’s scores omitted from the analysis. In
this analysis, no significant score differences were found (p /
0.85) across the group, nor was there a significant interaction
between subject and scheme factors (p /0.34).
and across subscales. The conventional device was preferred by
two subjects (S34, S35) for the subscale aversiveness, two
subjects (S36, S37) for the subscale ease of communication, and
one subject (S37) for the subscales of background noise and
reverberation. The frequency-compression device was preferred
by two subjects (S34, S38) for the subscale background noise,
one subject (S34) for the subscale reverberation, and one
subject (S36) for the subscale of aversiveness. Globally, higher
scores for the conventional device are shown for four of the six
subjects. Of the remaining two subjects, S38 showed a higher
score for the frequency-compression device whereas S39
showed no difference in scores between the two devices.
Discussion
Figure 4. Mean signal-to-noise ratios (SNRs) obtained by five
subjects for a target score of 50% words correct in sentences.
Unfilled bars show SNRs obtained with the conventional device
(CD), and filled bars show SNRs obtained with frequencycompression (FrC). SNRs averaged across subjects are shown in
the top pair of bars, with error bars indicating one standard
deviation. Statistical significance is shown by asterisk symbols: *
0.01 B/pB/0.05, *** pB/0.001.
Perceptual performance for the seven subjects who participated
in the study was similar, on average, between the frequencycompression scheme and the conventional hearing aids for words
and consonants presented at a moderate level in quiet conditions.
For individual subjects, no significant differences were found
between the conventional device and the frequency-compression
device for the CNC words test in quiet. For the test of
consonants, one subject (S37) showed a significant score
improvement, and two subjects (S32, S34) showed a significantly
lower score when listening with the frequency-compression
device. It is possible that the frequency-compression device
may have resulted in some perceptual improvements for certain
sounds at the expense of others, resulting in subjects’ overall
percentage correct score remaining unaffected. To investigate this
further, confusion matrices were constructed from the subjects’
consonant responses with both the conventional hearing aids and
the experimental device. These matrices are shown in Table 2.
The frequency compression resulted in some improvements.
Certain fricative phonemes such as /sh/ and /j/ were correctly
identified by subjects more often with the frequency compression
device than with the conventional device. Unfortunately, the
experimental scheme also resulted in some reductions in scores.
For example, the phoneme /g/ was frequently mistaken for /z/.
The recognition of /s/ was also reduced. Interestingly, some
fricative sounds such as /sh/, /z/, and /v/ were selected more often
as responses by subjects when using the frequency-compression
device. It is possible that subjects were hearing more fricative-like
sounds with the experimental scheme but were unable to identify
these correctly on every presentation.
Perhaps the experimental processing may have resulted in
improved intelligibility of speech with further training. Subjects’
experience with the frequency-compression device was limited.
Three of the subjects were not hearing-aid users at the time of
participation, and these subjects wore the frequency-compression device infrequently. The logged time that they wore the
device was on average 2.2 hours per week.
Subjectively, scores were higher in the APHAB questionnaire
for the frequency-compression scheme for only one participant
(S38). Unfortunately, results did not correspond well to the
objective speech test results. For example, S38 showed a score
decrease for both the word and consonant test when listening
with the device. In contrast, S37 showed higher scores for the
conventional device in the APHAB questionnaire, yet obtained
some improvements in speech perception in quiet when using the
frequency-compression device.
626
International Journal of Audiology, Volume 45 Number 11
Subjective assessment
For the APHAB questionnaire, the score each subject obtained
with the conventional device for each subscale was subtracted
from the score the subject obtained with the frequency-compression device. These values are shown for six subjects in Figure 5.
One subject (S32) did not return the APHAB questionnaire.
Positive values indicate a preference for frequency compression, whereas negative values indicate preference for the
conventional device. Preference ratings varied across subjects
Figure 5. Preference scores from the APHAB questionnaire provided by six subjects who participated in the study. For each subject,
the subscale Ease of Communication (EC) is shown by the filled black columns, the subscale Background Noise (BN) is shown by the
filled grey hatched columns, the subscale Reverberation (RV) is shown by the unfilled columns with horizontal lines, the subscale
Aversiveness (AV) is shown by the filled grey columns, and Global scores are shown by the unfilled columns. As shown on the vertical
axis, negative values indicate a preference for the conventional device (CD), whereas positive values indicate a preference for the
frequency-compression device (FrC).
The speech test results are consistent with those previously
reported by McDermott & Dean (2000) for six subjects with
steeply sloping losses who showed no significant differences
when frequency shifting was enabled compared to when it was
disabled. In that study, the tests included recognition of
monosyllabic words with proportional frequency shifting in
which all frequencies were lowered by a factor of 0.6.
The present results are inconsistent with those reported
previously for an identical signal processing scheme (Simpson
et al, 2005). Differences in the audiogram configurations across
studies may account for the discrepancy. For example, S35 had
the best high-frequency hearing thresholds of the group and was
also the one subject who showed significant improvements when
using the frequency-compression device for tests in noise.
Subjects who participated in the Simpson et al (2005) study
had, on average, some degree of low-frequency hearing loss
sloping to a severe to profound high-frequency loss for
frequencies above 2000 Hz. As noted, the current subjects had,
on average, near-normal hearing thresholds in the low frequencies, with a sharp drop in thresholds to severe-profound levels for
frequencies above 1000 Hz. It is likely that these subjects had
high-frequency dead regions based on the severity of their
hearing thresholds (Moore, 2001). Therefore, much of the
compressed signal used in this study may have been presented
in the dead-region range of hearing for these subjects. This may
have caused part of the signal to be inaudible, or audible but not
helpful for discriminating the speech signals used in the tests.
The resulting lack of useful additional information may account
for the frequency-compression scheme showing no overall
benefit for speech perception in these experiments.
There may be a limit on how much information can be
‘squeezed’ into a listener’s range of usable hearing. In Simpson et al
(2005), the cut-off frequencies were set between 1600 2500 Hz
Frequency-compression outcomes in
listeners with steeply sloping audiograms
because this was close to the lowest frequency at which the
hearing thresholds were greater (worse) than 90 dB HL. In the
current study, a lower cut-off frequency range (1250 1600 Hz)
was selected partly because of the different audiogram configurations of these subjects. One possible effect of the steeper
audiogram slopes is that some components of the frequencycompressed signal were inaudible, and that cut-off frequencies
below 1250 Hz could have resulted in an improvement in speech
perception. However, due to the non-linearity of the compression, a lower cut-off frequency would likely have resulted in a
decrease in sound quality. Any increase in high-frequency
information might have been at the expense of distorting
lower-frequency signals, such as vowel sounds. As described
earlier, four of the subjects found the sound quality of the
frequency-compression device unacceptable with a cut-off frequency of 1250 Hz. In these cases, the cut-off frequency was
increased to 1600 Hz.
Based on these observations, audiogram configuration is
suggested to be an important factor when recommending
frequency compression. The frequency at which the hearing
loss becomes severe as well as the steepness of the slope of the
audiogram may be important. Listeners with a steeply sloping
hearing loss can expect to receive limited benefit from the
experimental frequency-compression scheme tested in this study.
However, based on previously reported results (Simpson et al,
2005), some listeners with a more moderate slope may obtain
some additional cues about high-frequency signals when wearing
the frequency-compression device.
Conclusions
The performance of a frequency-compression device was evaluated by comparing the speech understanding abilities of seven
Simpson/Hersbach/McDermott
627
628
Table 2. Confusion matrices for the consonant test
International Journal of Audiology, Volume 45 Number 11
Stimulus
CD response
p
t
k
b
d
g
m
n
s
sh
z
f
ch
j
th
v
p
t
k
b
19
15
5
4
2
22
22
26
d
g
23
33
4
1
10
7
29
40
s
sh
z
f
2
Ch
5
6
1
7
20
j
th
v
7
11
1
4
4
12 121 11
1
4
4
10
1
3
25
1
82
1
1
27
4
6
5
2
18
1
11
68
1
1
2
15
13
2
61
5
1
24
6
40
5
3
1
6
1
7
61
1
1
2
4
11
18
13
2
3
1
38
30
1
3
1
1
N
1
9
1
m
FrC response
10
1
39
1
13
3
26
3
25
2
6
6
2
28
15
19
p
t
k
b
d
g
m
n
s
sh
z
f
ch
j
th
v
p
t
k
b
14
10
3
1
2
5
26
28
32
d
g
4
4
4
3
15
22
24
2
32
sh
5
9
21
33
32
1
4
1
2
s
z
f
ch
j
th
v
1
1
1
n
2
2
10
3
1
m
8
7
10
6
1
8
1
16 114 13
2
59
6
56
65
4
6
14
1
17
10
3
3
12
1
8
22
1
3
8
9
1
39
9
51
1
1
23
2
4
2
13
71
1
2
5
4
1
16
6
5
2
3
6
20
3
6
29
30
25
3
7
2
1
3
3
1
10
1
29
Note: Correct responses for each phoneme are shown in bold type diagonally. The total number of times each phoneme was selected as a response is shown in bold type at the bottom
of each matrix.
4
19
26
hearing-impaired listeners with steeply sloping hearing losses, in
both quiet and noisy conditions. The results can be summarized
as follows.
1. Use of the frequency-compression scheme provided similar
recognition of monosyllabic words and consonants as
conventional hearing aids, on average.
2. Improvements when listening to sentences in noise were
found for one subject when using the frequency-compression
device.
3. The results of the APHAB questionnaire showed higher
global scores for four of the six subjects for the conventional
device over the frequency-compression scheme.
Acknowledgements
The authors are grateful for the financial support of the
Commonwealth of Australia through the Cooperative Research
Centre for Cochlear Implant and Hearing Aid Innovation. The
Garnett Passe and Rodney Williams Memorial Foundation
provided financial support for the third author. Phonak Hearing
Systems provided the conventional hearing instruments used
in the study. The comments of Robert Cowan, Prof. Brian
Moore, and two anonymous reviewers on a previous version
of this manuscript are greatly appreciated. Cathy Sucher,
Rodney Millard, Justin Zakis, and WaiKong Lai are among
the many colleagues we’d like to thank for contributing to this
work. The study would not have been possible without the
subjects who generously gave their time to participate in the
experiments.
References
Baer, T., Moore, B.C. & Kluk, K. 2002. Effects of low pass filtering on
the intelligibility of speech in noise for people with and without dead
regions at high frequencies. J Acoust Soc Am , 112, 1133 44.
Bennett, D.S. & Byers, V.W. 1967. Increased intelligibility in the
hypacusic by slow-play frequency transposition. J Aud Res, 7,
107 118.
Boothroyd A., Hanin L. & Hnath T. 1985. Internal report RCI 10,
Speech and Hearing Sciences Research Centre, City University of
New York.
Byrne, D., Dillon, H., Ching, T., Katsch, R. & Keidser, G. 2001. NALNL1 procedure for fitting nonlinear hearing aids: Characteristics
and comparisons with other procedures. J Am Acad Audiol , 12, 37 51.
Byrne, D., Dillon, H., Tran, K., et al. 1994. An international comparison
of long-term average speech spectra. J Acoust Soc Am , 96, 2108 2120.
Cox, R.M. & Alexander, G.C. 1995. The abbreviated profile of hearing
aid benefit. Ear Hear, 16, 176 86.
Daniloff, R.G., Shriner, T.H. & Zemlin, W.R. 1968. Intelligibility of
vowels altered in duration and frequency. J Acoust Soc Am , 44, 700 7.
Davis-Penn, W. & Ross, M. 1993. Pediatric experiences with frequency
transposing. Hearing Instruments, 44, 26 32.
Fu, Q.J. & Shannon, R.V. 1998. Effects of amplitude nonlinearity on
phoneme recognition by cochlear implant users and normal-hearing
listeners. J Acoust Soc Am , 104, 2570 7.
Hochberg, Y. & Tamhane, A.C. 1987. Multiple Comparison Procedures.
Hobeken, NJ: Wiley.
Frequency-compression outcomes in
listeners with steeply sloping audiograms
Hogan, C.A. & Turner, C.W. 1998. High-frequency audibility: Benefits
for hearing-impaired listeners. J Acoust Soc Am , 104, 432 41.
Johansson, B. 1961. A new coding amplifier system for the severely hard
of hearing. Proceedings of the 3rd International Congress on
Acoustics. Stuttgart, 1959 , 2, 655 657.
Ling, D. 1968. Three experiments on frequency transposition. American
Annals of the Deaf , 113, 283 294.
McDermott, H.J. & Dean, M.R. 2000. Speech perception with steeply
sloping hearing loss: Effects of frequency transposition. Br J Audiol ,
34, 353 61.
McDermott, H.J., Dorkos, V.P., Dean, M.R. & Ching, T.Y. 1999.
Improvements in speech perception with use of the AVR TranSonic
frequency-transposing hearing aid. J Speech Lang Hear Res, 42,
1323 35.
McDermott, H.J. & Knight, M.R. 2001. Preliminary results with the
AVR ImpaCt frequency-transposing hearing aid. J Am Acad Audiol ,
12, 121 7.
Moore, B.C. 2001. Dead regions in the cochlea: Diagnosis, perceptual
consequences, and implications for the fitting of hearing aids. Trends
Amplif, 5, 1 34.
Moore, B.C., Huss, M., Vickers, D.A., Glasberg, B.R. & Alcantara, J.I.
2000. A test for the diagnosis of dead regions in the cochlea. Br J
Audiol , 34, 205 24.
Moore, B.C.J. & Glasberg, B.R. 1997. A model of loudness perception
applied to cochlear hearing loss. Aud Neurosci , 3, 289 311.
Murray, N. & Byrne, D. 1986. Performance of hearing-impaired and
normal-hearing listeners with various high frequency cut-offs in
hearing aids. Aust J Audiol , 8, 21 28.
Neary, T. 1989. Static, dynamic, and relational properties in vowel
perception. J Acoust Soc Am , 85, 2088 2113.
Parent, T.C., Chmiel, R. & Jerger, J. 1998. Comparison of performance
with frequency transposition hearing aids and conventional hearing
aids. J Am Acad Audiol , 9, 67 77.
Peterson, G.E. & Lehiste, I. 1962. Revised CNC lists for auditory tests. J
Speech Hear Disord , 27, 62 70.
Rankovic, C.M. 1991. An application of the articulation index to hearing
aid fitting. J Speech Hear Res, 34, 391 402.
Reed, C.M., Hicks, B.L., Braida, L.D. & Durlach, N.I. 1983. Discrimination of speech processed by low-pass filtering and pitch-invariant
frequency lowering. J Acoust Soc Am , 74, 409 19.
Reed, C.M., Schultz, K.I., Braida, L.D. & Durlach, N.I. 1985.
Discrimination and identification of frequency-lowered speech in
listeners with high-frequency hearing impairment. J Acoust Soc Am ,
78, 2139 41.
Rosenhouse, J. 1990. A new transposition device for the deaf. The
Hearing Journal , 43, 20 25.
Simpson, A., Hersbach, A.A. & McDermott, H.J. 2005. Improvements
in speech perception with an experimental nonlinear frequencycompression hearing device. Int J Audiol , 44, 281 292.
Thornton, A.R. & Abbas, P.J. 1980. Low-frequency hearing loss:
Perception of filtered speech, psychophysical tuning curves, and
masking. J Acoust Soc Am , 67, 638 43.
Turner, C.W. & Hurtig, R.R. 1999. Proportional frequency-compression
of speech for listeners with sensorineural hearing loss. J Acoust Soc
Am , 106, 877 86.
Velmans, M. 1974. The design of speech recoding devices for the deaf. Br
J Audiol , 8, 1 5.
Velmans, M. & Marcuson, M. 1983. The acceptability of spectrumpreserving and spectrum-destroying transposition to severely hearing-impaired listeners. Br J Audiol , 17, 17 26.
Vickers, D.A., Baer, T. & Moore, B.C.J. 2001. Effects of lowpass filtering
on speech intelligibility for listeners with dead regions at high
frequencies. Br J Audiol , 35, 148 149.
Zakis, J.A., McDermott, H.J. & Fisher, M. 2004. An advanced sound
processing system for hearing aid research. Iran J Audiol , 3, 75 78.
Simpson/Hersbach/McDermott
629
Minerva Access is the Institutional Repository of The University of Melbourne
Author/s:
Simpson, Andrea
Title:
Improving high-frequency audibility for hearing-impaired listeners using a cochlear implant or
frequency-compression aid
Date:
2007-05
Citation:
Simpson, A. (2007). Improving high-frequency audibility for hearing-impaired listeners using
a cochlear implant or frequency-compression aid. PhD thesis, Department of Otolaryngology,
The University of Melbourne.
Publication Status:
Unpublished
Persistent Link:
http://hdl.handle.net/11343/39236
File Description:
Improving high-frequency audibility for hearing-impaired listeners using a cochlear implant or
frequency-compression aid
Terms and Conditions:
Terms and Conditions: Copyright in works deposited in Minerva Access is retained by the
copyright owner. The work may not be altered without permission from the copyright owner.
Readers may only download, print and save electronic copies of whole works for their own
personal non-commercial use. Any use that exceeds these limits requires permission from
the copyright owner. Attribution is essential when quoting or paraphrasing from these works.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement