Widexpress 8: Senso: Audiological background

Widexpress 8: Senso: Audiological background
8
April 1997
Carl Ludvigsen
Senso: Audiological background
Introduction
The SENSO hearing instrument
from Widex represents a technological breakthrough in the hearing
aid industry given the utilization of digital signal representation throughout the entire signal
path from microphone to output
transducer. By developing digital technology to a level applicable for use in hearing aids, most
of the limitations previously imposed on the complexity of hearing aid circuitry have been eliminated. This degree of freedom
has been utilized to address a
number of unsolved problems experienced by hearing aid users.
SENSO utilizes this digital technology in three different areas.
First, to increase speech intelligibility in difficult listening environments by incorporating a
number of digital signal processing algorithms. Secondly, to alleviate some of the problems traditionally known to bother hearing aid users: acoustic feedback
and internal noise from the microphone, and thirdly, to optimize the fitting for the individual patient by providing the audiologist/hearing aid professional
with a versatile fitting tool.
This article treats the first of
these issues by reviewing the
audiological background which
constitutes the basis for the
SENSO signal processing. The
audiological principles take their
starting point in the functional
differences between the normal
and the impaired ear.
1
in the inner ear. It is well known
that this transformation is made
by the interaction of the two
types of sensory cells: the inner
hair cells (IHC) and the outer
hair cells (OHC). These hair cells
are arranged in four rows along
the basilar membrane: one row of
inner hair cells and three rows of
outer hair cells.
Normal Hearing
Normal hearing is based upon
the conversion of an incoming
sound into a nerve code in the
sensory cells of the Organ of Corti
2
1
Fig. 1: Cross section of the Organ of Corti in the inner ear. Inner (1) and outer (2) hair cells
are indicated with arrows.
1. This article is a slightly expanded version of “Basic Amplification rationale of a DSP-hearing Instrument,” Hearing Review,
March 1997.
Some psycho-acoustic
consequences of a loss
of hair cells
1
2
A
E
E
A
E
A
E
A
E
A
E
A
Fig. 2: Inner (1) and outer (2) hair cell with afferent (A) and efferent (E) nerve cells, see also
Fig. 1.
During the past decade it has become evident that the inner hair
cells are the factual sensory
cells. Thus, it is the inner hair
cells which, during acoustomechanical excitation, transform
the mechanical movement of the
sound stimulus into a complicated pattern of nerve impulses.
These nerve impulses are transmitted to the auditory cortex via
a structured network of nerve fibers and synapses.
The outer hair cells function, in
many aspects, like a servo-control. This servo-control mechanism has the effect that a low
level or soft sound is able to
evoke sufficiently large vibrations of the basilar membrane,
which in turn will excite the inner hair cells. This specialized
function of inner and outer hair
cells is reflected in several observations. E.g. in a close-up picture of the hair cells (figure2), it
is observed that the nerve cells
which innervate the inner hair
cells are predominantly of the
afferent type (marked with A).
These afferent fibers transmit
impulses from the sensory cell to
the brain. The outer hair cells,
on the other hand, are innervated by efferent nerve fibers
(marked with E). These efferent
fibers transmit nerve pulses
from the central part of the hearing organ to the sensory cell.
Moreover, chemical analyses have
shown that the outer hair cells
contain bio-chemical compounds
like actin, which are similar to
those found in muscle tissue.
When outer hair cells are stimulated, they have the potential of
changing their length and in so
doing, they enhance the vibrations of the basilar membrane.
The outer hair cells react almost
instantaneously to stimulation
and they are able to change their
shape with a very fast rate, i.e.
more than 20.000 times per second.
The function of the outer cells is
non-linear in the sense that at
very weak sound levels there is a
significant effect of outer hair
cell function, whereas more intense signals generate almost no
effect in function.
Hearing Loss
A sensorineural hearing loss may
have a number of causes. Typically, a loss of hair cells is the
primary causative factor for a
sensorineural hearing impairment.
A loss of outer hair cells will create a malfunction of the characteristic servo mechanism, causing weak sounds to be inaudible,
while more powerful sounds
(which in normal hearing do not
involve the contribution of the
outer hair cells) will be perceived with normal loudness.
This well-known recruitment phenomenon is thus characterized by
an increased threshold of hearing, HTL, without a similar shift
in the threshold where the signal becomes uncomfortably loud,
UCL. In this way, the effect of a
loss of outer hair cells can be
thought of as a loss of compression.
Inner hair cells are less vulnerable than the outer hair cells.
However, when inner hair cells
are damaged, one consequence
of this is a loss of selectivity at
all input levels. Reduced frequency selectivity, and the corresponding spread of masking,
reduces the ability to communicate in background noise.
Recruitment
The vast majority of hearing
losses include a loss of outer hair
cells and, as a consequence, will
involve loudness recruitment.
(It should be noted that recruitment is normally not associated
with retrocochlear lesions nor
the presence of a conductive
hearing impairment).
Loudness perception by
hearing impaired persons
The loudness with which a sound
is perceived is stated in a subjectively defined scale: the Sone2
scale. For sounds well above the
hearing threshold, i.e., at medium to high levels, experiments
have shown that, in normal hearing, the loudness will double as
the level increases with 10 dB
(i.e., at high suprathreshold levels, an increase of 10 dB will result in a doubling of the per-
2. The Sone scale is related to the physical dB scale by defining that the loudness 1 Sone corresponds to the average loudness
perceived by normal hearing listeners when listening to a 1000 Hz pure tone at a level of 40 dB SPL. The Sone scale is a so-called
ratio scale, and n Sones denotes the loudness of a sound which appears n times as loud as a sound with the loudness of 1 Sone.
ceived loudness). In figure 3, recruitment is exemplified by showing the perceived loudness (in
Sones) of a 1000 Hz tone for a normal hearing individual versus an
individual presenting a 40 dB
hearing impairment.
dB SPL
THRESHOLD OF DISCOMFORT
120
AUDIBLE AREA
AUDIBLE AREA
100
100
MUSIC
MUSIC
80
80
SPEECH
60
SPEECH
60
40
40
20
20
IMPAIRED THRESHOLD OF HEARING
NORMAL THRESHOLD
OF HEARING
0
10
Perceived Loudness in Sones
dB SPL
THRESHOLD OF DISCOMFORT
120
20
50
100 200
0
500 1000 2000 5000 10000 Hz
10
20
50
100 200
500 1000 2000 5000 10000 Hz
10
Fig. 4: The auditory range for a normal hearing and a person with a cochlear hearing loss.
4
2
cruitment due to a sensorineural
hearing impairment generally
have a reduced dynamic range as
compared to normal hearing individuals.
1
Normal
0.4
0.2
Hearing impaired
0.1
0
50
Input Level in dB SPL
100
Fig. 3: Loudness versus the level of a 1000
Hz pure tone. Curves are for a normal
hearing person and a person with a 40 dB
hearing loss.
In figure 3 it can be seen that a 1
kHz tone of 40 dB SPL generates
a perceived loudness of 1 Sone for
a normal hearing person. However, the tone must be 15 dB more
intense, i.e. 55 dB SPL, in order
to give the 40 dB hearing impaired person a similar loudness
perception. Thus, at this intensity level, a 15 dB insertion gain
will normalize the loudness of
this tone. At higher input levels,
however, the loudness perception tends to normalize for the
hearing impaired person when
compared to the normal hearing
individual. For example, in figure 3, at an input level of 75 dB
SPL, both the normal hearing
and the hearing impaired obtain
a loudness of 10 Sones. Thus, at
this level, no additional gain is
required at 1000 Hz. Figure 3
may, therefore, be used to calculate the insertion gain as a function of the input sound pressure
level which is required to obtain
‘loudness normalization’ for a
1000 Hz tone.
This is illustrated in figure 4
where the dynamic range for a
normal versus a hearing impaired individual is illustrated.
It is evident that for the hearing
impaired individual, the dynamic range is reduced relative
to that of the normal hearing individual.
The differences in loudness perception between the normal versus the hearing impaired individual are used in the development of the algorithms which
determine the amplifier characteristics of the SENSO.
Masking
Masking designates the phenomenon whereby the presence
of one sound may cause another
sound to be inaudible or to reduce its loudness, the latter being known as partial masking.
Masking may take place in frequency and in time. Thus, a tone
with a certain frequency may
mask a tone of another frequency
present at the same time, or up
until 200 ms after the sound has
ended.
Recruitment also reduces the
range of sound which can be detected by the recruiting ear. The
area between the threshold of
audibility and the threshold of
discomfort defines the dynamic
range of hearing. It is well
known from the literature that
individuals suffering from re-
level of test tone
80
dB
LM=90dB
60
40
70
20
50
100 200
500 1K 2K
frequency of test tone
The excessive spread of masking
is closely related to other functional deficits associated with
hair cell loss. These include impaired frequency resolution which
denotes that the ability to separate frequency components which
appear at the same time is reduced. These symptoms are likely
to involve listening problems for
hearing impaired individuals
when listening to speech in background noise.
Speech Understanding
50
30
0
20
Spread of Masking and
frequency resolution
Figure 5 shows the result of a
masking experiment. In the experiment is determined the presentation level required of tones
of different frequencies in order
to be heard when a simultaneous 1000 Hz masker tone is present. When the frequency of the
tone is close to 1000 Hz, the
masking is significant while no
masking occurs at remote frequencies. Figure 5 also indicates
that the masking curves are asymmetrical and that the masking effects are greatest toward
the higher frequencies. This phenomenon is referred to as an upward spread of masking. It is
widely acknowledged that many
hearing impaired individuals experience excessive upward spread
of masking, which has the consequence that intense low frequency
sounds such as noise may mask
weak speech components to a
higher degree than in normal
hearing.
5K
10KHz
Fig. 5: Masked thresholds with 1000
Hz pure tone maskers at levels ranging
from 20 to 90 dB SPL.
Speech in quiet
Speech understanding in quiet
seldom causes any significant
problems for the hearing impaired, sometimes even in the
0
-10
-20
-30
-40
-50
70
80
90
100
Level of Speech in dB SPL
Fig. 6: Intelligibility of speech in noise as
a function of overall level. Lower bold
curve is for hearing impaired listeners,
upper thin curve is for normal hearing
listeners (adapted from Studebaker,
Sherbecoe & McDaniel 1995).
Speech in noise
It is well known that most consonant sounds are much weaker in
acoustic energy than vowel sounds.
Normal speech intelligibility requires the perception of weaker,
high frequency consonants as
well as more powerful low frequency speech components which
are higher in acoustic energy. In
a situation where low frequency
background noise is present, the
ability to perceive these high frequency components will be disproportionately more difficult for
the hearing impaired individual
with impaired frequency resolution than for the normal hearing
individual. The excessive masking, experienced by the hearing
impaired individual, will add to
the elevated threshold and result in a loss of “speech understanding” or “distinctness” of
speech due to the loss of consonantal speech segments.
proved speech understanding.
When the speech level is raised
above a certain level, a more or
less pronounced drop in intelligibility may be the consequence.
This effect is more pronounced
for hearing impaired listeners
than for listeners with normal
hearing. Figure 6 illustrates the
negative effect on speech intelligibility for hearing impaired
and normal hearing individuals
when the overall level of speech
and noise is increased beyond a
certain limit, and that the negative effect is more pronounced
for hearing impaired listeners
than for normal hearing individuals.
Overamplification
An increase in the speech level
does not always lead to im-
Influence of noise type
It is a well known experience
from studies of multiple program
Thus, it can be seen that the
hearing aid user experiences benefit from amplification only to a
certain limit, and that overamplification results in negative effects.
hearing aids that hearing impaired listeners prefer a different frequency response in different noise environments. Keidser
(1995) has studied this phenomenon by comparing the preference
among three frequency characteristics and found (see figure 7)
that the frequency response defined by the NAL fitting rule
was preferred when listening to
speech in a background of babble
noise while a steeper slope with
less bass and more treble than
NAL was preferred in a background of traffic noise and a
shallower response with more
bass and less treble was preferred in a quiet background. In
our own research, we have observed similar effects provided
the reduction of gain did not affect the audibility of the speech
signal. These observations have
been utilized in the formulation
of the SENSO signal processing.
aaa
aaa
6
FLATTER
5
4
Average score
Relative score in %
absence of a hearing instrument. This is primarily due to
the presence of superfluous information or cues i.e., visual, contextual, structural, situational,
phonemic etc., denoted as redundancy. In a quiet environment,
the redundancy may compensate adequately for the loss of
acoustic information caused by
the hearing impairment. Severe
problems arise when additional
acoustic information becomes
unavailable, e.g. due to masking
by background noise.
3
2
1
0
aa
aa
aa
aa
aa
aa
aa
aa
aa
aa
aa
aa
aa
aa
aa
aa
aa
aa
aa
aa
Quiet
NAL
STEEPER
aa
aa
aa
aa
aa
aa
aa
aa
aa
aa
aa
Babble
aa
aa
aa
Traffic
Fig. 7: Preference of frequency response in
different listening environments (adapted from Keidser, 1995).
The SENSO Hearing Instrument
Basic Amplification
Template of SENSO
this summation of loudness over
frequency and time.
The signal processing of SENSO
aims to compensate for the negative effects which result from a
sensorineural hearing impairment. More specifically, SENSO's
signal processing algorithms,
which determine its amplification as a function of time and frequency, aim to compensate for
loudness recruitment and excessive masking. This is accomplished by developing digital signal processing algorithms which
operate according to two underlying, continuously operating principles, Loudness Mapping and
Noise Reduction. According to the
first principle, the presentation
level and frequency balance will
be continuously adjusted to compensate for the long term effects
of recruitment. According to the
second principle, noise reduction, the sound image will be further shaped in order to minimize
the effect of a detrimental listening situation e.g. with heavy
background noise. In different
noise backgrounds, this latter
compensation gives a change in
the frequency response which
corresponds to the change in preferred frequency response in different listening environments.
The first attempts to compensate
for loudness recruitment in the
early 70es, used compression with
very short time constants (Villchur 1973). This fast regulation
may seem the logical choice when
compensating the lost function of
the fast acting outer hair cells.
However, experiments have shown
that a fast acting compression has
a tendency to create distortion and
to enhance noise during speech
pauses. This so-called pumping
effect was perceived as annoying by
most users. With regard to speech
intelligibility, the fast acting compression did not provide quite as
good speech intelligibility in most
situations as did a slow acting compression and our own research
showed that a slow regulation
was preferred by the majority of
hearing aid users.
The Loudness Mapping
principle
The goal of Loudness Mapping is
to present complex signals like
speech to the hearing aid user in
such a way that their different
components are perceived with
the same loudness as a normal
hearing person perceives them.
Thus, it is necessary over the entire frequency range, to transform the wide dynamic range of
speech (in different environments) to fit within the narrow
dynamic range of the impaired
listener (i.e., restoring a sense of
“natural loudness”). Loudness of
a complex sound depends, in a
complicated way, on its spectral
content and its temporal course,
and most attempts to restore natural loudness have disregarded
plex tone signals to the user. The
advantage of this method is that
the individual acoustic properties of the user’s ear and earmould are included in the threshold values and these are more
suitable for hearing aid fitting
than normal audiogram values
obtained with a transducer different from the one used in the
hearing aid. The three thresholds obtained in this way are
used to determine the target input/output (I/O) curves for each
of the three frequency bands.
This is accomplished according
to a Loudness Mapping scheme,
which is based on the measured
differences in loudness perception for normal hearing versus
hearing impaired listeners reported by Pascoe (1988).
dB HL
150
140
130
120
Thus we concluded that the loudness mapping should predominantly include an adjustment of
level in each frequency band according to the long term level of
the speech signal. This slow adjustment may be implemented in
a way which provides audibility
of the speech signal without blurring its temporal structure and
without creating severe pro- cessing artifacts like pumping or
distortion.
Consequently, the loudness mapping strategy implemented in
SENSO is primarily based on the
long term properties of the incoming signal. Since it is known
that in fast varying acoustic environments a slow regulation may
have some drawbacks, separate
algorithms have been developed
to minimize these side effects.
Practical implementation
The starting point for SENSO’s
loudness mapping is an in-situ
determination of the hearing
threshold in each of SENSO’s
three frequency bands. This is
obtained by letting the hearing
aid generate and present com-
UCL
110
100
90
80
70
60
MCL
50
500 Hz
40
1000 Hz
30
HTL
20
10
2000 Hz
4000 Hz
0
0
10 20 30 40 50 60 70 80 90 100 110 120
dB HL
Fig. 8: The uncomfortable level, UCL,
and the most comfortable level, MCL, for
several frequencies recorded as a function
of HTL, from Pascoe (1988).
Pascoe's (1988) data explored the
auditory dynamic range and recorded MCL and UCL judgments
in reference to hearing threshold
levels for a large group of hearing
impaired individuals. The data
revealed that the relations between MCL or UCL and the
hearing loss, were independent
of frequency (when measuring
the input level in dB HL, see figure 8) i.e. the mean MCL and
UCL judgments for each hearing
threshold level did not show a
significant effect of frequency.
We found that Pascoe’s relation
between average HTL and UCL
was well approximated by a second order polynomial and we utilize this relation in the SENSO
fitting algorithm for estimating
UCL thresholds from the individual HTL values. Pascoe’s mean
UCL values are slightly lower
(approximately 10 dB) than
those found in most studies (for a
recent discussion see Elberling
and Nielsen, 1993), but, by utilizing Pascoe's data, we are
choosing more conservative values which provide a margin to
fine tune for individual deviations from the mean UCL. This
possibility to fine tune calculated UCL values is of particular
significance, clinically, given the
range or variation of individual
UCL thresholds observed for a
given audiological configuration.
dB HL
140
120
100
80
60
40
20
0
0
20
40
60
80
100
120
dB HL
dB HL
140
120
100
80
60
40
20
Studies of loudness perception of
hearing impaired listeners have
shown that recruitment takes
place, in particular, at levels just
above the hearing threshold (e.g.
Hellman, 1990) and is manifested by a steepening of the
loudness function. This observation has been included in a calculation template which enables
the calculation of a probable
loudness function at any given
hearing loss. This template is used
in the calculation of SENSO's
Loudness Mapping function and
is illustrated in figures 9 and 10.
0
0
20
40
60
80
100
120
dB HL
Fig. 9: Loudness template and corresponding input/output characteristic.
dB HL
HTL:60
140
120
100
80
60
40
20
0
This loudness calculation template is likely to predict loudness
for pure tones and narrow band
noise, which are not representative of daily listening environments. Typically, environmental
sound is comprised of speech (or
music) possibly in the presence of
background noise or reverberation. Such input signals are
characterized by having a broad
band spectrum and by fluctuating considerably over time. They
are consequently neither stationary nor narrow band. For
such signals, the calculation of
loudness is far more complicated
than when considering stationary narrow band signals. Recent
research, however, has indicated
that no substantial improvement
is likely to be achieved by introducing more sophisticated loudness models when compensating
takes place, may therefore be set
as low as the individual earmould or shell allows. When using a fast regulation, a kneepoint lower than approximately
45 dB SPL is rejected by a majority of the users. Note the apparent paradox, that just above
threshold where the action of
the outer hair cells and thereby
the recruitment is at its maximum, most users do not accept a
fast recruitment compensation.
With the SENSO a kneepoint as
low as 15-20 dB SPL has been
positively received.
0
20
40
60
80
100
120
dB HL
Fig. 10: I/O curve with gain and output
limitation.
for loudness recruitment. Our
own research has shown that
the combination of the slow acting, statistically based type
loudness mapping and the noise
reduction approach used in the
SENSO has been positively
judged by practically all test persons, also when asked specifically about the loudness of the
amplified signal. A special feature of the slow acting compression is that users generally accept a very low compression
kneepoint, and the fitting tools
allow for an individual setting of
the kneepoints in all bands. The
lower kneepoint, representing the
level above which compression
Temporal aspects
In most situations, especially
those with a fairly continuous
background noise, we have found
that a slow acting regulation was
preferred. This is consistent with
a number of recent studies (e.g.
Neuman et al. 1994). In some
situations, however, where impulse noise is present, a faster
regulation was generally preferred. Consequently, in the
SENSO we have incorporated a
complex regulation algorithm
which combines both fast and
slow acting regulation to address these environmental differences. As long as the listening
environment is stable, the release time will be very long and
the regulation will be slow. In
this situation, the SENSO will
act as a perfectly linear device,
with corresponding low distortion. However, if the environment changes, if the noise level
increases, or if a sudden, loud
impulse noise arises, the regulation times will automatically become faster. In such situations
SENSO’s mode of operation is
highly nonlinear.
Noise reduction
A second principle, denoted
“Noise reduction”, modifies the
SENSO Loudness Mapping
scheme according to a statistical
analysis of the input signal, carried out independently within
three separate bands. The noise
reduction aims at improving the
hearing impaired listener’s possibilities of listening to speech in
severe background noise. In such
noise, speech is produced with a
higher than normal effort, which
gives rise to a shift in spectrum
and a higher overall level. If a
band contains intense noise, this
will inevitably cause masking.
In such a band, the speech level
is typically well above threshold
and a gain reduction may significantly reduce masking without corrupting the available
speech information. In order to
utilize this principle, it is necessary to ascertain what is speech
and what is noise. This is of
course a difficult task bearing
in mind that background noise
often consists of multi-talker
speech. We approached that problem by studying the statistical
properties of various signals. It
appeared that the speech from a
single talker differed substantially from the statistical properties of various types of noise including multi- talker babble, see
figure 11.
25
Speech
Frequency in %
20
15
10
5
0
-60
-50
-40
-30
-20
-10
0
10
-10
0
10
Level in dB / rms
25
Noise
Frequency in %
20
15
10
5
0
-60
-50
-40
-30
-20
Level in dB / rms
Fig. 11: Level distributions of speech and
noise. Short term levels are in dB re. long
term rms value.
Using digital technology it proved
possible to implement a comprehensive statistical analysis of the
level distribution in each band.
This continuously running analysis of the incoming signal uses
time windows with durations of
up to 30 seconds. The analyses
provide running estimates of the
composition of speech and noise
in each band. As an example, if
the statistical analysis classifies
the input signal as a speech signal in intense background noise,
then the amplifier characteristic
is modified to reduce the loudness
of the noise (and the speech).
This modification takes into account the fact that the speech
signal should remain well above
threshold. In “Noise reduction”,
therefore, the “sound image” is
adjusted in order to optimize
speech intelligibility by minimizing masking effects. This
statistical analysis of the input
signal is a continuously ongoing
facet of SENSO's digital signal
processing.
Speech
30 dB
Speech
Noise
30 dB
Noise
Hearing Threshold
Hearing Threshold
Fig. 12: Principle of the noise reduction
strategy in SENSO: If no noise is detected, the speech signal is reproduced directly according to the loudness mapping
template.
Throughout this statistical analysis, a running update of the parameters which control the
SENSO sound reproduction takes
place. This principle is illustrated
in figure 12. When listening to
speech in a noise-free environment, the hearing instrument
characteristics (the control parameters) are calculated according to the loudness mapping
template without any modifications. As such, speech will be
audible in its full dynamic range
across each of the three frequency bands. However, should
speech be present in the context of
background noise, or if noise
alone is present, the calculation
scheme is modified in order to
minimize masking effects through
gain reduction.
Aside from the audiological benefits provided by this speech enhancement/noise reduction algorithm, a secondary benefit to the
user is a reduction of fatigue in
the presence of severe background noise. This may explain
the increased duration of hearing aid use observed in preliminary field studies.
Summary and
conclusions
In summary, the SENSO sound
reproduction is determined according to in-situ thresholds
measured by means of the hearing aid itself. The sound reproduction is basically determined
by the long term statistical properties of the different band levels,
but additional fast operating
functions have been included in
order to handle situations with
fast varying listening situations.
Moreover, a noise reduction paradigm is added on top of the loudness mapping. The noise reduction is based on a continuous statistical analysis of the levels in
each band which result in an estimate of the speech and noise
level in this band and possibly in
a relative reduction of the band
amplification. In this way, the
sound reproduction of SENSO
not only depends on the hearing
loss configuration and the spectrum of the sound environment,
but also on the nature of the
acoustic environments in which
the instrument is utilized.
References
Bülow, M (1996). “Field test evaluation of
SENSO by Widex,” Widexpress 7.
Elberling C. and Nielsen C. (1993) “The
dynamics of speech and the auditory dynamic range in sensorineural hearing
impairment,” Proc. XVth Danavox symposium Edited by J. Beilin and G.R.
Jensen. 99-133.
Hellman, R.P. and Meiselman, C.H.
(1990). “Loudness relations for individuals and groups in normal and impaired
hearing,”
J.
Acoust.
Soc.
Am.
88,2596-2606.
Keidser, G. (1995). “The relationship between listening conditions and alternative amplification schemes for multiple
memory hearing aids,” Ear and Hear.
16,6,575-586.
Neuman, A.C., Bakke, M.H., Mackersie,
C., Hellman, S. and Levitt, H. (1995).
“Effect of release time in compression
hearing aids: Paired-comparison judgments of quality”, J. Acoust. Soc. Am.
98,6,3182-3187.
Studebaker G.A, Sherbecoe R.L, and
McDaniel (1995). “The effect of higher
than normal signal and noise levels on
the speech recognition performance of
hearing impaired listeners,” Presented at
the 1st Hearing Aid Research and Development Conference, Bethesda USA.
Villchur, E. (1973). “Signal processing to
improve speech intelligibility in perceptive deafness,” J. Acoust. Soc. Am.
53,6,1646-1657.
Pascoe D.P (1988). “Clinical measurements of the auditory dynamic range and
their relation to formulas for hearing aid
gain,” Proc 13th Danavox Symposium.
Edited by Janne Hartvig Jensen.
129-151.
(P!W0P!5208!1012T)
Printed by FB / 3-97
P W0P 5208 101
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