lzz rzq- 1?.`) gm`.
US 2006026293 8A1
(19) United States
(12) Patent Application Publication (10) Pub. No.: US 2006/0262938 A1
(43) Pub. Date:
Gauger, JR. et al.
(54)
ADAPTED AUDIO RESPONSE
(52)
(76) Inventors: Daniel M. Gauger JR., Cambridge,
MA (US); Christopher B. Ickler,
Sudbury, MA (US); Nathan Hanagami,
Framingham, MA (US); Edwin C.
Johnson JR., Ashland, MA (US)
Correspondence Address:
MINNEAPOLIS, MN 55440-1022 (US)
(57)
ABSTRACT
Adapting an audio response addresses perceptual effects of
an interfering signal, such as of a residual ambient noise or
other interference in an earpiece of a headphone. In one
aspect, an input audio signal is presented substantially
an earpiece of a headset, and use the measured level in
conjunction With the level of an input audio signal to
(21) Appl. No.:
11/131,913
(22) Filed:
May 18, 2005
determine compression characteristics Without requiring
separation of an interfering signal present in the monitored
acoustic signal from a component related to the input audio
signal. In another aspect, presentation characteristics of an
input audio signal are determined to reduce distraction from
Publication Classi?cation
Int. Cl.
H04R 29/00
.............................................................. .. 381/56
unmodi?ed When it is at levels substantially above the
interfering signal and is compressed When at or beloW the
level of the interfering signal. The approach can make use of
a measured level of an acoustic signal, for example, Within
FISH & RICHARDSON PC
PO. BOX 1022
(51)
US. Cl.
NOV. 23, 2006
an interfering signal, such as from a background conversa
tion.
(2006.01)
N01 se saw-c9- ?lo
__.
l
.
5441M‘. Ms‘!
lzz
rzq»
9-341
T
mwcz
p114
Miguel-0on1
mpu
,
1?.‘)
gm‘.
1:. '
hubuué
R
1.:
Patent Application Publication Nov. 23, 2006 Sheet 1 0f 7
US 2006/0262938 A1
Nolse sourct lac
13°
3
Audio '3‘:
source ?
‘
Yo‘
c‘
y
e.
camel
I
12.0
"
__.
I
am». was»?
I22.
‘
I29
'
nan):
Woe»
s‘
j
I“00+ I31
HEN!“
n3
I26)
-
HASKuuq
WLE
“HQ
mfg-pphonq.
l'npu
FIG. 1
Jar
1
I #3311“
HoOuLE
d
Patent Application Publication Nov. 23, 2006 Sheet 2 0f 7
US 2006/0262938 A1
100
— - - -Audio under aan'pieoe if no noise
—— — Residual ambient noise if audlo turned off
9° -
Audio adjusted by NAUC
SdMELiacPeroBvpLtihenlc,
20
-90
I
l
I
I
I
I
I
-80
-7O
-60
~50
4O
-3O
-20
Audio Slgnal Input Level, dBV
FIG. 2A
~10
Patent Application Publication Nov. 23, 2006 Sheet 3 0f 7
m.23:E1u09-2O5
US 2006/0262938 A1
w
L.
\ Gbp
-20
BPc
-15
SNSR, 65
BP:
FIG. 2B
Patent Application Publication Nov. 23, 2006 Sheet 4 0f 7
0
,.-..-.-.‘.............
2:1; , . . . . . . .
m
....
. . . . . . . .
-15
U
2
0,
-2o
~25
-30
FIG. 2C
......
US 2006/0262938 A1
a-f
Patent Application Publication Nov. 23, 2006 Sheet 5 0f 7
a
§€3M+
“W
US 2006/0262938 A1
5118-34
_ a — n _ ~ '__
g?m’ismué‘s- “
33,’ ‘ ~11
|
‘g
l ',r_- -
Il
.
@323
—
— _ —
l
(MM
1
mo
l
‘
' I
'
l
‘ SPF
343,’
\ ah
l
all»)
“
‘
F
8’
\z»
{
“4
W
imahone
a
'
(as)
‘f 2
W -r
33'
3Q?
‘
l
|
33':
l
<48)
'
“ML/non:
an} ‘
:
‘
‘
31.x _ _ __ _ -2
21!
'52
I
l ‘as’:
awaom
1\ 12-5
‘
_ _~
1
am
Lil“
L
_ _ — ~ g _
M
122.
--—‘
Patent Application Publication Nov. 23, 2006 Sheet 6 0f 7
123
US 2006/0262938 A1
‘no
I25
1%
(A,
4/’!
Ema-‘P2: .
‘124
‘a:
I112
a":
420
.-
Cam
I!" i
M3 ‘M
‘"51
0573558
M6 b
[1‘!
""
IR
Mew/hm
in; u ‘'
FIG. ‘I’
Patent Application Publication Nov. 23, 2006 Sheet 7 0f 7
US 2006/0262938 A1
Nov. 23, 2006
US 2006/0262938 A1
ADAPTED AUDIO RESPONSE
BACKGROUND
[0001] This invention relates to adaptation of an audio
response based on noise or other interfering ambient signals.
[0002] When one listens to music, voice, or other audio
over headphones, one is often seeking a private experience.
Using the headphones presents the audio in a fashion that
does not disturb others in one’s vicinity and hopefully
prevents sounds in one’s environment (i.e., ambient noise
such as conversation, background noise from airplanes or
trains, etc.) from interfering With one’s enjoyment of the
audio.
[0003] Ambient noise can intrude on the quiet passages
unless one listens to the audio at a suf?ciently high volume,
Which may make subsequent loud passages uncomfortable
or potentially dangerous. Using closed-back, noise-reduc
ing, and especially active-noise-reducing (ANR) head
phones can help by reducing the level of ambient noise at the
ear. Even using such noise reduction, the available dynamic
range betWeen the maximum level one Would like to hear
and the residual ambient noise level after reduction by the
headphone is often less than the inherent dynamic range of
the input audio. This is particularly true With Wide dynamic
range symphonic music. One recourse is to repeatedly adjust
the volume control in order to enjoy all passages of the
music. Similarly, in situations in Which one Wishes to use the
[0008] In another aspect, in general, a method for process
ing an audio signal includes receiving the audio signal and
monitoring an acoustic signal that includes components of
an interfering signal and the audio signal. A processed audio
signal is generated. This includes compressing the audio
signal at a ?rst compression ratio When the audio signal is at
a ?rst level determined from the monitored acoustic signal
and compressing the audio signal at a second compression
ratio When the audio signal is above a second level deter
mined from the monitored acoustic signal. The ?rst level is
loWer than the second level and the ?rst compression ratio
is at least three times greater than the second compression
ratio.
[0009] Aspects can include one or more of the folloWing
features.
[0010] Generating the processed audio signal further
includes selecting a compression ratio according to a rela
tionship betWeen a level of the audio signal and a level of the
acoustic signal.
[0011] The relationship betWeen the level of the audio
signal and the level of the acoustic signal is determined
Without separating the components of the interfering signal
and the audio signal.
[0012] Processed the audio signal reduces a masking e?fect
related to the interfering signal. For example, the masking
e?fect related to the interfering signal can include at least one
of reducing an intelligibility of the interfering signal, reduc
ing a distraction by the interfering signal, and partially
music as a background to cognitive activities, the user may
adjust the volume so that the input music or other signal
masks distractions present in the ambient noise While not
intruding too much onto one’s attention.
masking the interfering signal.
[0013] Generating the processed audio signal includes
[0004] Approaches to adapting a speech signal for pre
adjusting at least one of a gain and a compression of the
audio signal according to a masking e?fect related to the
sentation in the presence of noise have made use of com
pression With the goal of achieving good intelligibility for
the speech. Some such approaches compress the speech
using a single compressor ratio, Where said slope is com
puted from the available dynamic range determined from an
estimate of the noise level and a maximum desired sound
level (e.g., a loudness discomfort level).
SUMMARY
[0005] In one aspect, in general, a method for adapting an
audio response addresses perceptual effects of an interfering
signal, such as of a residual ambient noise or other inter
interfering signal and to the audio signal.
[0014] The second compression ratio can take on a value
including approximately one to one, and a value less than
tWo to one.
[0015] The ?rst compression ratio can take on a value
including a value that is at least three to one, and a value that
is at least ?ve to one.
[0016]
The second compression ratio can be applied When
a level of the audio signal is at least 10 dB above a level of
the interfering signal.
ference in an earpiece of a headphone. An input audio signal
is presented substantially unmodi?ed When it is at levels
[0017]
substantially above the interfering signal and is compressed
[0018] The acoustic signal is monitored in the earpiece.
[0019] A source of the interfering signal is outside of the
When at or beloW the level of the interfering signal.
[0006]
In another aspect, in general, a method for adapting
The processed audio signal is transmitted to an
earpiece.
earpiece.
an audio response makes use of a measured level of an
acoustic signal, for example, Within an earpiece of a headset,
and uses the measured level in conjunction With the level of
an input audio signal to determine compression character
istics Without requiring separation of an interfering signal
present in the monitored acoustic signal from a component
related to the input audio signal.
[0007]
In another aspect, in general, a method for adapting
an audio response adjusts presentation characteristics of an
[0020] The acoustic signal includes at least some compo
nent of the audio signal.
[0021] Monitoring the acoustic signal outside an earpiece.
[0022] Applying active noise reduction according to the
acoustic signal.
[0023] Determining a time-varying relationship betWeen a
level of the audio signal and a level of the acoustic signal.
[0024] Generating the processed audio signal includes
input audio signal, for example for presentation in a headset
earpiece, to reduce distraction from an interfering signal,
varying a gain of the audio signal over time according to the
such as from a background conversation.
time-varying relationship.
Nov. 23, 2006
US 2006/0262938 A1
[0025] Generating the processed audio signal comprises
varying a degree of compression of the audio signal over
time according to the time-varying relationship.
[0026] The audio signal is expanded When the audio signal
is beloW a threshold level.
[0027] In another aspect, in general, a method for audio
processing involves receiving an audio signal, and monitor
ing an acoustic signal that includes components related to
both the audio signal and an interfering signal. Arelationship
betWeen a level of the audio signal and a level of the acoustic
signal is determined. Determining this relationship is per
[0041] In another aspect, in general, a method for audio
processing includes receiving an audio signal, and monitor
ing a level of an acoustic signal that includes components of
an interfering signal and the received audio signal. The
audio signal is processed. The processing includes com
pressing the audio signal When the level of the acoustic
signal is beloW a ?rst level and maintaining the audio signal
substantially unmodi?ed When the level of the acoustic
signal is above a second level.
[0042]
Aspects can include one or more of the folloWing:
[0043] Compressing the audio signal When the acoustic
formed Without separating the components related to the
signal is beloW a ?rst level includes applying a compression
audio signal and the interfering signal. The audio signal is
ratio that is at least three to one. The compression ratio can
also be at least ?ve to one.
processed according to the relationship to mitigate a per
ceptual effect of the interfering signal producing a processed
audio signal.
[0028] Aspects can include one or more of the folloWing
features.
[0029] Determining the relationship betWeen the level of
the audio signal and the level of the acoustic signal is
performed Without reconstructing the interfering signal.
[0030] The processed audio signal is presented in an
earpiece.
[0044] Maintaining the audio signal substantially unmodi
?ed includes passing the audio signal Without substantial
compression. For example, a compression ratio can be
applied that is approximately one to one over a range of
levels of the acoustic signal When a level of the audio signal
is at least 3 dB above a level of the interfering signal. As
another example, such a one-to-one compression action is
applied When the level of audio signal is at least 10 dB above
the level of the interfering signal.
[0045] A level of the interfering signal is determined based
[0031] Monitoring the acoustic signal includes monitoring
an acoustic signal in the earpiece.
[0032] Determining the relationship between the audio
signal and the acoustic signal comprises determining a
relative level of the audio signal and the acoustic signal.
[0033] An active noise reduction approach is applied to
the monitored acoustic signal.
on a level of the acoustic signal.
[0046] In another aspect, in general, a method for process
ing an audio signal includes receiving an audio signal and
monitoring a level of an acoustic signal that is related to the
audio signal. The audio signal is processed by compressing
the audio signal at a compression ratio of at least three to one
When the acoustic signal is beloW a ?rst level and compress
ing the audio signal at a compression ratio of substantially
[0034] The perceptual effect of the interfering signal
one to one When the acoustic signal is above a second level.
includes one or more of a masking by the interfering signal
The second level can be greater than the ?rst level.
and a distraction by the interfering signal.
[0035]
Mitigating the perceptual e?fect includes one or
more of masking the interfering signal using the audio signal
and reducing an intelligibility measure of the interfering
signal.
[0036] Determining the relationship betWeen the level of
the audio signal and the level of the acoustic signal includes
determining a time-varying relationship betWeen those lev
[0047] In another aspect, in general, a method for reducing
a perceptual effect of an interfering signal includes receiving
an audio signal and monitoring an acoustic signal that
includes components of the audio signal and the interfering
signal. A level of the audio signal is controlled according to
a level of the acoustic signal to reduce the perceptual effect
of the interfering signal, thereby creating a processed audio
signal.
els.
[0048]
[0037] Processing the audio signal includes varying a gain
[0049] Controlling the level of the audio signal includes
of the audio signal over time according to the time-varying
relationship, or varying a degree of compression of the audio
signal over time according to the time-varying relationship.
audio signal according to a masking effect of the interfering
signal on the audio signal.
[0038] Processing the audio signal comprises amplifying
[0050]
portions of the audio signal according to a relative level of
the audio signal and the acoustic signal. For example, a
greater gain is applied to loW level portions of the audio
earpiece.
signal relative to the gain applied to high level portions of
the audio signal.
[0039] The processed audio signal is substantially the
Aspects can include one or more of the folloWing:
adjusting at least one of a gain and a compression of the
The processed audio signal is transmitted to an
[0051] Monitoring the acoustic signal includes monitoring
the acoustic signal in the earpiece.
[0052] A source of the interfering signal is outside of the
earpiece.
same as the audio signal When the audio signal is above a
threshold level.
[0053] Active noise reduction is applied according to the
[0040] Processing the audio signal includes expanding the
[0054] In another aspect, in general, an audio processing
audio signal When the audio signal is beloW a threshold
level.
microphone for monitoring an acoustic signal, the acoustic
acoustic signal.
system includes an input for receiving an audio signal and a
Nov. 23, 2006
US 2006/0262938 A1
signal including components related to the audio signal and
[0070]
an interfering signal. A tracking circuit determines a rela
tionship betWeen a level of the audio signal and a level of the
processing includes processing a desired signal, monitoring
acoustic signal Without separating the components related to
the audio signal and the interfering signal. A compressor
circuit processes the audio signal according to the relation
ship to mitigate a perceptual effect of the interfering signal.
audio signal and an interfering signal, and determining a
relationship betWeen the desired audio signal and the acous
tic signal Without requiring separation of the desired signal
and the interfering signal. Processing the desired signal
[0055]
Aspects can include one or more of the folloWing:
perceptual effect of the interfering signal.
[0056]
The compressor circuit compresses the audio sig
[0071] In another aspect, in general, an audio processing
nal When the acoustic signal is beloW a ?rst level and
maintains the audio signal substantially unmodi?ed When
the acoustic signal is above a second level. The second level
can be greater than the ?rst level.
[0057]
The compressor circuit compresses the audio sig
nal at a compression ratio of at least three to one When the
In one aspect, in general, a method for audio
a signal that includes components related to the desired
includes using the determined relationship to mitigate a
system includes a compression module, Which accepts an
audio signal input and a microphone input. The compression
module includes circuitry to monitor the microphone input,
circuitry to determine a relationship betWeen the audio
signal input and the microphone signal Without requiring
separation of the audio signal input from the microphone
input, and circuitry to process the audio signal input using
acoustic signal is beloW a ?rst level and compresses the
audio signal at a compression ratio of substantially one to
one When the acoustic signal is above a second level.
the determined relationship to mitigate a perceptual effect of
[0058] The system includes an earpiece, the microphone
being external to the earpiece.
[0072] Aspects can include one or more of the folloWing
features.
[0059] The acoustic signal monitored by the microphone
includes a minimal component of the audio signal.
[0073] An earpiece, including a microphone inside the
earpiece that provides the microphone input, and a driver
coupled for presenting the processed audio input. The com
an interfering signal present in the microphone input.
[0060] The system includes an earpiece containing the
pression module can be housed in the earpiece.
microphone and a driver.
[0074] A masking module that accepts an audio signal
[0061] At least one of the tracking circuit and the com
pressor circuit is in the earpiece.
input and the microphone input. The masking module
includes circuitry for processing the audio signal input
[0062] A masking module accepts an audio signal input
and the microphone input, the masking module including
circuitry for processing the audio signal input according to
a level of microphone input, including controlling a level of
the audio signal input to reduce a perceptual effect of an
interfering signal present in the microphone input.
according to a level of microphone input, including control
ling a level of the audio signal input to reduce a perceptual
effect of an interfering signal present in the microphone
input.
[0075]
[0076]
[0063]
A selector selectively enables at least one of the
compression circuit and the masking module.
A selector to selectively enable one or the com
pression module and the masking module.
Embodiments can have one or more of the folloW
ing advantages.
includes a ?rst input for receiving an audio signal and a
[0077] Estimation of the noise level in the absence of
audio does not necessarily have to be computed alloWing
adaptation of the audio signal based on measures of the
second input for receiving a microphone signal that includes
audio level as Well as level of the audio plus residual
components related to the audio signal and an interfering
signal. A correlator processes the audio signal according to
ambient noise under the earpiece For example, direct deter
mination of the gain and/or compression ratio to be applied
based on a SNSR value (ratio of signal to noise plus signal)
[0064] In another aspect, in general, a masking module
a level of the microphone signal and a level of a modi?ed
audio signal. A level of the modi?ed audio signal is con
trolled to mitigate a perceptual effect of the interfering
signal.
[0065]
Aspects can include one or more of the folloWing:
[0066]
A control circuit that controls the level of the
modi?ed audio signal.
[0067] The control circuit adjusts the level of the modi?ed
audio signal such that the output of the correlator is main
tained substantially equal to a threshold value.
[0068]
The control circuit includes a smoothing ?lter, such
measured in an earpiece of a headphone is enabled. This can
avoid a relatively computationally expensive signal process
ing, Which is desirable a portable, battery-poWered system.
[0078] Determination of the gain from the SNSR by
comparing the audio signal input to the total signal (repro
duced audio plus residual noise) at a microphone under the
earpiece can offer several advantages. As a result of the
relationship betWeen SNR and SNSR, a tWo-segment piece
Wise linear relationship describing gain as a function of
SNSR results in a smooth transition from uncompressed to
highly compressed audio.
responsive to an output of the correlator and an output of a
user controllable correlation target.
[0079] A user is able to choose Whether he or she Would
like to experience that music in the presence of noise in one
of tWo different manners. One manner, termed “upWard
[0069] A bandpass ?lter coupled to each of the micro
range of the music to be heard by the user in the presence of
phone signal and the modi?ed audio signal.
noise While preserving the inherent dynamic qualities of the
as an integrator, an output of the smoothing ?lter being
compression,” has the goal of alloWing the full dynamic
Nov. 23, 2006
US 2006/0262938 A1
music. Rather than applying a simple compression of the
audio, Which could affect the dynamic qualities of relatively
loud passages, the audio that is quiet enough to be masked
by the noise is adapted, but When the music signal is
substantially louder than the noise, substantially no com
unit 110 (e.g., through the design ofearpiece 112 and ear pad
114 ) and optionally using an active noise reduction system
embedded in the headphone unit. The audio signal input 131
is processed in the headphone unit in a signal processor 120
and a driver output signal 127 is passed from the signal
pression is applied thereby preserving the dynamic qualities.
processor 120 to a driver 116, Which produces the acoustic
The other manner, termed “auto-masking,” has the goal of
using the audio to prevent the user being distracted by
realiZation of the audio signal input. The user perceives this
acoustic realiZation in the presence of an interfering signal,
speci?cally in the presence of the attenuated ambient noise.
The signal processor may alternatively be located external to
aspects of the noise, primarily conversations of nearby
people.
[0080] In another aspect, in general, softWare includes
instructions for execution on a digital processor to perform
all the steps of any of the methods described above. The
softWare can be embodied on a machine-readable medium.
[0081] In another aspect, in general, a system for audio
processing includes means for receiving an audio signal, and
means for monitoring an acoustic signal that includes com
ponents related to both the audio signal and an interfering
signal. The system also includes means for determining a
relationship betWeen a level of the audio signal and a level
of the acoustic signal. Determining this relationship is
performed Without separating the components related to the
audio signal and the interfering signal. The system includes
means for processing the audio signal according to the
relationship to mitigate a perceptual effect of the interfering
earpiece 112.
[0092] A number of transformations of the audio signal
input 131 that are performed by the signal processor 120 are
based on psychoacoustic principles. These principles
include masking effects, such as masking of a desired audio
signal by residual ambient noise or masking of residual
ambient noise by an audio signal that is being presented
through the headphones. Another principle relates to a
degree of intelligibility of speech, such as distracting con
versation, that is presented in conjunction With a desired
signal, such as an audio signal being presented through the
headphones. In various con?gurations and parameter set
tings, the headphone unit adjusts the audio level and/or
compression of a desired audio signal to mitigate the effect
of masking by ambient noise and/or adjusts the level of a
signal producing a processed audio signal.
desired signal to mask ambient noise or to make ambient
conversation less distracting. In some versions, the user can
[0082]
select betWeen a number of different settings, for example,
to choose betWeen a mode in Which the headphones mitigate
Other features and advantages of the invention are
apparent from the following description, and from the
claims.
ambient noise and a mode that makes ambient conversation
less distracting.
DESCRIPTION OF DRAWINGS
[0083]
FIG. 1 is an overall block diagram of a headphone
[0093] The signal processor 120 makes use of an input
from a microphone 118 that monitors the sound (e.g., sound
audio system.
pressure level) inside the earpiece that is actually presented
[0084]
to the user’s ear. This microphone input therefore includes
components of both the acoustic realiZation of the audio
FIG. 2A is a graph relating an audio signal input
level and an output audio level.
[0085] FIG. 2B is a graph of compression module gain
versus signal-to-(noise+signal) ratio (SNSR).
[0086] FIG. 2C is a graph relating the signal-to-noise ratio
(SNR) to the SNSR.
[0087]
ule.
FIG. 3 is a block diagram of a compression mod
[0088]
FIG. 4 is a block diagram of a masking module
sion module 122 performs a level compression based on the
noise level so that quiet audio passages are better perceived
by the user. A masking module 124 performs gain control
and/or level compression based on the noise level so the
ambient noise is less easily perceived by the user. A noise
[0089] FIG. 5 is a block diagram of a noise reduction
module.
DESCRIPTION
1 System OvervieW (FIG. 1)
[0090]
signal input and the attenuated (or residual) ambient noise.
[0094] The signal processor 120 performs a series of
transformations on the audio signal input 131. A compres
Referring to FIG. 1, an audio system 100 includes
a headphone unit 110 Worn by a user. The headphone unit
receives an audio signal input 131 from an audio source 130.
The audio source 130 includes a volume control 132 that can
be adjusted by the user. The user listens to an acoustic
realiZation of the audio signal input that is generated Within
reduction module performs an active noise reduction based
on a monitored sound level inside the earpiece. In alternative
versions of the system, only a subset of these modules is
used and/or is selectively enabled or disabled by the user.
2 UpWard compression (FIGS. 2A-C, 3)
[0095] For some modes of operation and/or parameter
settings, the compression module 122 provides level com
pression based on the noise level so that quiet passages are
better perceived by the user. The general approach imple
mented by the compression module 122 is to present por
tions of the audio signal input that are louder than the
ambient noise With little if any modi?cation While boosting
generates ambient acoustic noise. The ambient acoustic
quiet portions of the audio signal input that Would be
adversely affected by the ambient noise. This type of
approach is generally referred to beloW as “Noise Adapted
UpWard Compression (NAUC).” The result is a compres
sion of the overall dynamic range of the input audio signal,
noise is attenuated by the physical design of the headphone
Where the net amount of compression applied is a function
the earpiece.
[0091]
In general, a noise source 140, such as a source of
mechanical noise, people conversing in the background, etc.,
Nov. 23, 2006
US 2006/0262938 A1
both of the dynamic range of the input audio and the relative
NAUC modi?es the acoustic realiZation level at the ear due
level that the user Wishes to listen to compared to the
ambient noise level the user hears.
to the audio input. For input signals such that the uncom
pressed audio output level at the ear Would be Well beloW the
residual noise level (less than —80 dBV input as shoWn) the
signal processor provides a compressor module gain 235
[0096] NAUC is designed to account for masking caused
by residual ambient noise inside the earpiece. If this noise is
loud enough relative to an audio signal input, the noise can
that is as large as 25 dB.
render the audio signal inaudible. This effect is knoWn as
[0100] With moderate residual noise under the headphone
complete masking in the psycho-acoustic literature. The
signal-to-noise ratio (SNR) at Which complete masking
earpiece, if the user listens to audio that is substantially
louder than the residual noise, the audio is not appreciably
occurs is a function of various factors, including the signal
and noise spectra; a typical value is approximately —l5 dB
modi?ed by NAUC (this corresponds to the input signals
(i.e., the audio signal is 15 dB quieter than the residual
ambient noise). If the signal-to-noise ratio is greater than
above —45 dBV in FIG. 2A). If the user subsequently turns
the volume doWn so that the quieter portions of the music
approach or are less than the noise level, the compression
that needed for complete masking then partial masking is
module responds by amplifying those passages. The loWer
said to occur. Under conditions of partial masking, the
the audio signal input level relative to the residual noise
perceived loudness of the signal is reduced compared to
level, the more gain 235 is provided by the compression
When the masking noise is absent. In the range betWeen
module, until very loW audio levels are reached (less than
—80 dBV input as shoWn).
complete masking and no masking, the steepness of the
loudness function increases as compared to a noise-free
condition (i.e., a larger apparent change in signal loudness is
heard for a given change in objective signal level). When
listening to audio in the presence of residual ambient noise,
a user can set the volume control for the desired level of the
loudest passages of the music and the NAUC processing
applies a compression of the audio appropriate to the volume
setting. The NAUC approach provides audibility, and rea
sonably natural perception of the dynamics of the quieter
[0101] The gain characteristics of the NAUC compression
module as illustrated in FIG. 2A is not characterized by a
single compression ratio. If the user listens to music With a
limited dynamic range at a loud level relative to the residual
noise, the NAUC compression module reproduces the music
Without compression. As the audio volume setting is
decreased, the dynamic range is increasingly compressed. If
the parameters determining the shape of line 240 are suitably
passages in the presence of the noise.
chosen, the increasing compression With decreasing level
[0097] To illustrate the masking effect quantitatively,
by the noise. The result for the user is that the inherent
assume that the earpiece unit provides 20 dB of noise
reduction of ambient noise outside the headphones. For
example, While riding in an airliner With 80 dB SPL (Sound
Pressure Level) interior noise level, the attenuated ambient
dynamic qualities of the music, in the presence of the
residual noise and processed by the NAUC system, sound
noise at the ear is 80 dB minus 20 dB or 60 dB SPL. Assume
that the user is listening to symphonic music With a 60 dB
dynamic range and adjusts the volume control of the audio
source so that the crescendos are presented at the rather loud
level of 95 dB SPL. The quietest passages of the music Will
be at 95 dB minus 60 dB or 35 dB SPL. HoWever, the
attenuated ambient noise in this example is at 60 dB SPL,
and therefore the quietest passages are at an SNR of —25 dB,
compensates for the effects of partial masking of the audio
largely the same as When the music is listened to in the
absence of noise and Without compression.
[0102] For input signals such that the uncompressed audio
output level at the ear Would be Well beloW the residual noise
level, the compression module can continue to provide
increasing gain or, as shoWn for levels less than —80 dBV in
FIG. 2A, can preferably provide a doWnWard expansion
characteristic. In such a range, gain 238 decreases With
decreasing input level. DoWnWard expansion can be useful
Which is more than the typical threshold for complete
masking, so these quiet passages Will be completely masked.
In the NAUC approach, these quiet passages are ampli?ed
by ensuring that the self-noise ?oor of the audio source is not
(upWard compressing them) While not substantially chang
[0103] Referring to FIG. 3, the compression module 122
ampli?ed to the point that it becomes audible and objec
tionable.
ing the dynamics of the louder passages.
[0098] Referring to FIG. 2A, an example of a relationship
betWeen the level of the audio signal input Qi-axis 210) and
the level of the output acoustic realiZation of the audio signal
(Y-axis 212) for a particular level of ambient noise in the
earpiece. The dashed line 220 represents the residual ambi
ent noise level (60 dB SPL) in the earpiece. Note that this
ambient noise level is independent of the audio signal input
level. The output audio level that Would result in the earpiece
audio signal and residual ambient noise at the user’s ear.
Note that if the headphones include a noise reduction
of the signal processor 120 includes a signal/noise tracker
322, Which processes the audio signal input 131 and the
microphone input 119 to determine estimates related to the
audio signal input level and monitored audio microphone
level. In the present embodiment the monitoring microphone
is located inside an earpiece of the headphone; therefore the
microphone output includes components comprising the
as a function of the input signal, if it Were used in an
module 126, for example for active noise reduction (ANR),
environment With no ambient noise, is shoWn by the dash
dot line 230. This input-output relationship is linear (e.g., a
20 dB input level change causes a 20 dB output level
change) and re?ects an uncompressed gain for the head
one microphone 118 can be used for both ANR and NAUC
phone itself of 110 dB from the input (in dBV) to the output
signal/noise tracker 322.
(in dB SPL).
[0104] The signal/noise tracker 322 accepts the audio
signal input 131 and the microphone input 119. The micro
phone input 119 is applied to a multiplier 310 that multiplies
[0099]
In FIG. 2A, the solid curve 240 shoWs hoW the
compression module 122 that is con?gured to implement
signal processing. The input is processed through a gain/
compression processor 324 that applies gain and/or level
compression based on control information provided from the
Nov. 23, 2006
US 2006/0262938 Al
the input by a calibration factor to adjust the relative
sensitivity of the headphone system, and to make the micro
Whereas for high audio levels (SNR>0 dB) the SNSR
phone input after calibration and the audio signal input
essentially equal in level for typical audio signals in the
(equal levels for the residual ambient noise and the acoustic
realiZation of the audio signal) then SNSR=—3 dB. The
relationship betWeen SNSR and SNR (in dB) shoWn in FIG.
2C can be expressed mathematically (assuming no correla
tion betWeen the audio and noise) as:
absence of any substantial ambient noise. The tWo signals,
the audio signal input 131 and the calibrated microphone
input, are then passed through band-pass ?lters (BPF) 312
approaches a maximum value of 0 dB; for an SNR=0 dB
and 316, respectively, to limit the spectrum of each to a
desired range. In the present embodiment, the BPF blocks,
pass frequencies from 80 to 800 HZ. This bandWidth is
chosen because the response of a typical ANR headphone,
SNSR : lOlogl0[ 1 +loSNR/IO
loSNR/IO ]
from audio input to acoustic output in the earpiece, varies
less from Wearer to Wearer Within this range of frequencies
compared to other bandWidths. This frequency range also
encompasses most of the energy in typical audio signals.
Other BPF bandWidths could alternatively be used.
[0108] Referring again to FIG. 3, the SNSR and the output
[0105]
to apply to the audio signal. The gain/ compression processor
324 applies a time-varying gain to the audio that is deter
The signals from BPF blocks 312 and 316 are of
limited bandWidth and can be decimated or resampled to a
loWer sample rate in digital signal processing embodiments.
This alloWs the processing for blocks 314 and 318 and all
elements in gain/compression processor 324 except multi
plier 334 to be done at the decimated rate, reducing com
of the audio envelope detector 314 are passed to the gain/
compression processor 324 to determine the amount of gain
mined from the SNSR in a gain calculation block 330.
Referring to FIG. 2B, compressor gain 235 as a function of
SNSR 321 corresponds to the graph shoWn in FIG. 2A. This
gain is speci?ed according to a set of four parameters 328.
ment, the outputs of the BPF blocks are decimated to a 2.4
Speci?cally, in the present embodiment the gain is calcu
lated according to four parameters (BPZ, BPc, Gbp, and Sc)
putation and poWer consumption. In the present embodi
kHZ sample rate. Other rates, including full audio bandWidth
With different formulas being applied in three ranges of
may be used as Well.
SNSR as folloWs.
[0106]
[0109] For a range of SNSR>BPZ, the gain is 0 dB. In the
example shoWn in FIG. 2B, the breakpoint BPZ=—0.5 dB. A
SNSR of —0.5 dB corresponds to an SNR of approximately
10 dB (i.e., the signal level is Well above the noise masking
level), as indicated in FIG. 2C.
The outputs ofthe BPF blocks 312 and 316 are fed
into envelope detector 314 and 318, respectively. The func
tion of each envelope detector is to output a measure of the
time-varying level of its input signal. Each envelope detec
tor squares its input signal, time averages the squared signal,
and then applies a logarithm (10*log1O( )) function to
convert the averaged level to decibels. The tWo envelope
detectors have different averaging time constants for rising
and falling signal levels. In the present embodiment, the
envelope detector has a risetime of approximately 10 mil
liseconds and a falltime (release time) of approximately 5
seconds; other rise and fall time constants, including equal
values for risetime and falltime, can alternatively be used. A
rapid rise/sloW fall envelope detector is a common charac
teristic of audio dynamic range compressors, With the choice
of time constants being an can be important aspect of
minimiZing to minimiZe audible “pumping” of output signal
levels in response to changing dynamics of the input. In the
present system, referring to FIG. 2A, a fast risetime ensures
that, When the audio signal input level increases rapidly from
the partial or complete masking region (SNR<0 dB) to the
no masking region (SNR>0 dB) the compressor module gain
235 is rapidly reduced so the audio does not sound abnor
mally loud.
[0107] The outputs (in dB) of the envelope detectors 314
and 318 are subtracted at a difference element 320, audio
envelope minus microphone envelope, to produce an esti
mate of the audio signal-to-(noise+signal) ratio (SNSR) 321
present in the earpiece. If the calibration factor input to
multiplier 310 is properly set and With the headphone
operating on the head in a quiet environment (i.e., negligible
[0110] For SNSR=BPc (Where BPc<BPZ), the gain
applied is Gbp. For a range SNSR<BPc, a compression
slope of Sc on the gain as a function of SNSR is applied to
the input level. That is, for every 1 dB decrease in SNSR, the
gain increases by Sc dB. For audio levels Well beloW the
residual noise level (e.g., less than —10 dB SNR), SNSR
approximates quite closely the SNR, as shoWn in FIG. 2C.
The dependence of gain on SNSR thus results in a com
pression ratio of 1:(1-Sc). In the example in FIGS. 2B-C, the
BPc breakpoint is chosen to be at SNSR=—3 dB, Which
corresponds to an SNR of approximately 0 dB; this occurs
at an input level of —50 dBV in the FIG. 2A. In the example
of FIGS. 2A-B over a range of input levels the compression
slope Sc is chosen to be 0.8 Which corresponds to a
compression ratio of approximately 1:0.2, or 5:1. Over the
input range of —60 dBV (corresponding to —10 dB SNR)
doWn to —80 dBV FIG. 2A shoWs an approximately linear
increase in compressor module gain 235 as the input level
decreases.
[0111] In the intermediate region BPc<SNSR<BPZ, the
gain is linearly interpolated (as a function of SNSR) betWeen
a gain of 0 at SNSR=BPZ to gain of Gbp at SNSR=BPc as
shoWn in FIG. 2B. In the example, Gbp=3 dB. The range of
BPc<SNSR<BPZ corresponds to a range of audio signal
input level of approximately 10 dB, Which results in a range
of output level of 10 dB-3 dB=7 dB, appreciably less than
residual ambient noise) then typical audio signals should
result in equal envelope detector outputs, corresponding to
the 5:1 compression applied to loWer audio signal input
an SNSR of 0 dB. Referring to FIG. 2C, a graph of the
SNSR (Y-axis) as a function of the SNR (X-axis) shoWs that
in the presence of residual ambient noise, for loW audio
[0112] The gain calculation incorporating these param
levels (SNR<0 dB) the SNSR approximates the SNR
levels.
eters, implemented in 330 and outlined above, can be
expressed succinctly as folloWs:
Nov. 23, 2006
US 2006/0262938 A1
With the doWnWard expansion portion Will slide to the left on
the ?gure and the maximum gain provided by the compres
0
SNSR > BPZ
_ W
Gbp + (BPC - SNSR) * Sc
sion module Will decrease. If the zero reference level and
expansion slope are properly chosen, based on listening
experiments and the actual hardWare’s self-noise character
(SNSR - BPC)
SNSR < BPc
istics, the audibility of audio source or signal processor
self-noise is minimized. Other means of limiting gain for
loW audio signal input levels may also be used While
[0113] The equation above describes the compression
module gain 235 for audio inputs corresponding to
achieving the basic qualities of the NAUC system.
SNSR<BPz in terms of tWo segments, each of Which are
linear on SNSR and Which join at SNSR=BPc, as Well as the
segment of zero gain for SNSR>BPZ. Given the nature of the
rate limiting. It is presumed that the residual ambient noise
is in most cases nearly constant or sloWly varying; it is
undesirable to have the NAUC system suddenly amplify the
relationship betWeen SNSR and SNR, as illustrated in FIG.
2C, over the range —10 dB<SNR<10 dB, the pieceWise
linear relationship betWeen gain and SNSR (shoWn in FIG.
2B) results in a compressor gain 235 applied to the audio
input that smoothly transitions from the high compression
region (slope Sc, SNSR<—10 0 dB) to decrease toWard zero
compressor gain (slope 1, SNSR>10 dB), as shoWn in FIG.
2A. The effective compression that results in this region is
not characterized by a single slope as it is When SNSR<BPc.
[0114] The four parameters (BPz, BPc, Gbp and Sc) may
be chosen based on the psychoacoustic experiments on
partial masking but preferably are set based on comparative
listening to music both in the absence and presence of noise.
Chosen properly, these parameters ensure that the inherent
dynamic qualities of music are similar When it is listened to
over the headphones either in quiet or in the presence of
residual ambient noise. Other values than those presented in
the example above may be desirable. At least some choices
of the parameters provide approximate restoration of musi
cal dynamics in the presence of noise and, in particular, the
smooth transition from uncompressed audio for large signals
(much greater than 0 dB SNR) to highly compressed audio
for small signals (less than 0 dB SNR). Listening tests have
shoWn that compression ratios for small signals in excess of
3:1 and compression ratios for large signals substantially
[0117] In addition, gain limiter 332 incorporates gain sleW
audio in response to transient noises in one’s environment
such as results from accidentally tapping the earpiece or
coughing. To minimize this, the gain limiter in the present
embodiment limits the rate at Which gain can increase to a
rate of 20 dB/ second. No limit on the rate at Which gain can
decrease is applied so that the system reacts as determined
by gain calculation 330 to rapid increases in the audio signal
input level.
[0118]
The output of the gain limiter 332 is then converted
from decibels to a scale factor, passed through an anti
zipper-noise ?lter (to eliminate the audible effect of discrete
gain steps and then applied at a multiplier 334 to amplify the
audio signal input 131 producing an audio signal output 123
that is passed to the masking module 124.
[0119] A characteristic of at least some embodiments of
the system is the absence of a requirement to estimate the
noise level in the absence of audio. The gain is determined
from the SNSR (ratio of signal to noise plus signal) rather
than the SNR (ratio of signal to noise).
2.1 Alternatives
[0120] Alternatively, a microphone external to the head
phone’s earpiece(s) can be used to determine the noise level.
The signal level is adjusted for the noise attenuation of the
less than 2:1 (preferably 1:1) are desirable.
earpiece (passive and possibly ANR) and the sensitivity of
[0115]
The output of the gain calculation block 330 is fed
the headphone itself (gain from audio signal input level to
to a gain limiter 332 that limits that gain so that the gain is
not excessive for very loW audio signal input levels. An
effect of this gain limiter is to ensure that the gain is reduced
so that When the audio signal is loW or possibly absent (e.g.,
sound pressure level under the earpiece). Note that the
combined uncertainty in these factors can be signi?cant,
Which may result in a less accurate compensation of the
the audio source is turned on but not playing or during the
ever, there may be situations (e.g., in the case of open-back
silence betWeen musical tracks) the self-noise ?oor of the
headphones that provide little if any noise attenuation) in
Which placement of the microphone outside the earpiece
source itself is not ampli?ed to undesirable levels. In the
example shoWn in FIG. 2A, the gain limiter is determined
by ?rst computing a doWnWard expansion gain value equal
to the expansion slope times the difference, in dB, betWeen
the audio signal input level and a zero reference level. The
zero reference level corresponds to the audio signal input
level With no signal playing and for Which no compression
module gain is to be applied. The actual gain in dB to apply
to the audio signal is the minimum of the gain determined by
gain calculation 330 and this doWnWard expansion gain.
[0116]
In the example in FIG. 2A, the doWnWard expan
sion slope is 2:1 and the zero reference level is —95 dBV.
These values, along With the 60 dB SPL residual noise level
shoWn in FIG. 2A, alloW a maximum compressor module
effects of partial masking by the compressor module. HoW
outWeighs such potential uncertainty.
[0121] An SNSR based and under-earpiece-microphone
based compressor module, as described above, may also be
sensitive to hoW accurately the headphone and microphone
sensitivity is knoWn. An addition optional block can be
added to the block diagram of FIG. 3 to enable the system
to self-calibrate. This block Would take as inputs SNSR 321
and audio signal input envelope 315 and output the calibra
tion factor applied to multiplier 310. This optional block
adjusts the calibration factor sloWly to ensure that, When the
audio signal input envelope is large the SNSR is 0 dB.
Preferably the calibration factor is only updated to achieve
0 dB SNSR during intervals With large audio signal input
gain of approximately 25 dB (at audio signal input level of
envelope levels When said intervals folloW a short time after
—80 dBV). As the residual noise level is reduced, the point
at Which the high compression part of curve 240 intersects
intervals Where the audio level is substantially loWer While,
at the same time, SNSR Was moderate (in the vicinity of 0
Nov. 23, 2006
US 2006/0262938 A1
dB SNR). Assuming that the noise level is slowly changing,
this ensures that the calibration factor update only occurs
When the audio level signi?cantly exceeds the residual noise
level.
[0122] BPFs 312 and 316 may be designed so as to pass
a range of frequencies other than the 80 to 800 HZ range of
the present embodiment. Altemately, other ?lter character
istics than a band-pass response may be used to select the
gain limiting ?lters are non-causal, requiring the audio
signal input to be delayed an appropriate amount prior to
multiplier 334.
[0127] A simpler gain calculation 330 may be achieved by
setting the compressor gain, in dB, equal to a constant times
the negative of the SNSR. If the constant is Sc (G=—
SNSR’X‘Sc) then the resulting gain is very similar to that
shoWn in FIG. 2 A, With a maximum difference from the
portion of the audio input and monitored microphone signals
more complex, four parameter gain calculation described
from Which the levels are determined.
above of only 0.6 dB for Sc=0.8. Of course, the error using
[0123] Other implementations of the envelope detectors
different Gbp, Sc, BPc, and BPZ values. This simpler gain
calculation provides only one parameter determining the
314 and 318 can be used. For example, the envelope
detectors can operate on absolute values (i.e., signal mag
nitude) rather than squared values. This reduces the com
putational burden and computational dynamic range chal
lenges in ?xed-point DSP implementations. Also, logarithms
in bases other than base 10, other scale factors than 10 or 20
applied to the logarithm, or other non-linear functions may
be alternatively used to describe signal levels instead of
decibels. For example, truncated Taylor series expansions
may be used instead of the logarithm or poWer functions
(10") used in converting to and from the level units; these
can be computed over various ranges of values using coef
?cients from a lookup table that have been pre-computed.
This approach can be suf?ciently accurate While computa
tionally more e?icient than the logarithm or poWer function
in a ?xed-point DSP implementation.
[0124] Other envelope detection time constants than those
described above can be used. For example, equal values
could be used such as are used in speech envelope detectors
such a simpli?ed gain calculation Would be larger for
compression slope for SNSR<<0 dB. HoWever, no other
parameters are available to alloW ?ne tuning the operation of
the compression module in listening tests.
[0128] Alternatively, and though it could require addi
tional computational complexity, the gain calculation 330 as
a function of SNSR could use additional breakpoints or
alternative gain calculation arithmetic. The parameters used
in the envelope detection and gain calculation could also be
made to vary With audio or microphone level.
[0129] Alternatively, the upWard compression could be
done separately in different frequency bands, so as to better
approximate the psycho-acoustic characteristics of partial
masking at various levels or to mitigate the ampli?cation
into audibility of the audio source self-noise ?oor. If the
upWard compression is done in a multi-band fashion, it
could be desirable to have noise levels from loWer frequency
bands factor into the compression calculation at higher
frequencies so as to approximately compensate for the
(typically, 10 milliseconds). Alternatively, sloWer time con
psycho-acoustic effect of upWard spread of masking. This
stants can be used resulting in more of an automatic volume
could be done by (a) factoring in a fraction of the loWer
frequency SNSR or microphone level values in determining
the effective SNSR value in higher frequency bands used to
adjustment rather than compression characteristic in
response to the residual noise level. Another alternative is
for the envelope detectors to average by means of sleW rate
limits, either symmetric or asymmetric on the rise and fall,
rather than by means of rise and fall time constants created
by a ?lter With a feedback topology.
[0125]
The signal processing blocks shoWn in FIG. 3 can
be implemented in discrete time to occur at the sample rate
required for full audio bandWidth Without any decimation
after BPF blocks 312 and 316.
[0126] It is also desirable to have the microphone enve
lope detector 318 reject sudden transients such as are caused
by tapping an earpiece; the present embodiment incorporates
gain sleW rate limiting into gain limiter 332 for this purpose.
Rather than using identical time constants for audio and
microphone envelope detectors 314 and 318, different time
constants may also help mitigate the effect of transient
noises. The time constants used in the microphone level
detect 318 could also be made to vary as a function of the
outputs of the audio and microphone level detectors 314 and
318. For example, the microphone level detector could be
set to sloWly respond to changes except When a rapid rate of
change of the audio level is observed. Alternatively, more
sophisticated transient rejection can also be employed in the
gain limiter function such as using the median or mode
(most common value) of the level Within a moving WindoW.
Such alternate approaches can include variants of the
median or mode that respond differently to sudden increas
ing or decreasing gain transients. To be most effective such
compute compressor gain or (b) by making the bandpass
?lter prior to the microphone level estimate block have a
less-steep loWer frequency slope than the BPF prior to the
audio envelope detector block, thereby including some
loWer frequency noise energy in the SNSR determination for
that frequency band.
[0130]
It can also be desirable to have the system modify
the upWard compression characteristic during intervals
When no audio signal is present so that audio source or input
circuitry self-noise is not ampli?ed, becoming objection
able; the present embodiment includes an input audio level
dependent doWnWard expansion in gain limiter 332 to
achieve this. Multi-band operation can also achieve this.
Other approaches to achieve a loWering of gain during
intervals of very loW audio input level may also be used,
such as adjusting the upWard compression gain calculation
parameters (e.g., Gbp and Sc) as a function of input audio
level, microphone level or SNSR.
[0131]
Though reasons are given above stating Why an
SNSR-based compression determination is advantageous,
similar input-to-output characteristics as that represented by
line 240 in FIG. 2A can be achieved if an SNR estimate is
available. An estimate of the noise level could be determined
from the microphone level during intervals When the SNSR
is less than —l0 dB or a comparable threshold; this value
could be held ?xed in a memory register during intervals
When SNSR is greater than the threshold. The stored noise
Nov. 23, 2006
US 2006/0262938 A1
level estimate could then be used to determine an SNR value
as an input to a different gain computation. More sophisti
cated and computationally intensive parameter estimation or
adaptive ?lter techniques could be applied to estimate the
residual noise under the headphone earpiece, absent the
headphone audio, as Well. Also, signals derived Within the
noise reduction module can be used instead of the raW
microphone input 119. For example, the di?ference betWeen
the microphone input and the desired audio signal at the
di?ferencing element 530 (see FIG. 5) can be used. Alter
natively, a microphone external rather than internal to the
earpiece could be used to directly measure the noise and then
some calibration (representing the headphone’s noise
attenuation) applied to estimate the residual noise under the
earpiece. Given an SNR value obtained using any of the
above methods, the desired gain, including the uncom
pressed characteristic for SNR>>0 dB and highly com
pressed characteristic for SNR<<0 dB, can be computed
from a pieceWise linear or polynomial function.
liking for the task at hand. Examples of such selected audio
can be a steady noise (such as the masking noise sometimes
used to obscure conversation in open-plan of?ces), pleasant
natural sounds (such as recordings of a rainstorm or the
sounds near a forest stream), or quiet instrumental music.
[0137] A simple quantitative example can illustrate hoW
bene?cial this type of masking approach can be. Suppose the
user is Working in an open-plan of?ce With a background
noise level of 60 dB SPL resulting from the conversation of
one’s neighbors. If a headphone that provides 20 dB noise
reduction is donned, the resulting residual noise level of the
distracting conversation at the ear is 60 dB minus 20 dB, or
40 dB SPL. Although attenuated, this residual noise level
can be loud enough for a person With normal hearing to
easily understand Words and thus potentially be distracted.
HoWever, assuming that an SNR of —l0 dB (i.e., the ratio of
residual unattenuated conversation “signal” level to audio
input masking “noise” level) provides sufficient partial
masking so as to make the surrounding conversation unin
[0132] Compression of high-level audio signals could be
telligible (or at least not attention grabbing), then the user
added to ensure that the headphone does not produce pain
can listen to audio of the user’s choice at a level of 50 dB
fully loud, hearing damaging, or distorted audio levels.
SPL and obscure the distracting conversation. Thus, When
[0133] The parameters determining the upWard compres
Wearing such a system the user is immersed in 50 dB SPL
audio that the user prefers to Work by, as opposed to the 60
sion as a function of SNSR or SNR can be made user
dB SPL (i.e., 10 dB louder) background conversation that
adjustable, While maintaining the uncompressed character
may have distracted the user.
istic for SNR>>0 dB.
[0134] The embodiment described above implements
NAUC in a headphone. Noise adaptive upWard compression
can alternatively be applied in other situations, for example
in situations characterized by an approximately knoWn time
delay for propagation of output audio signal 123, through an
acoustic environment, to microphone signal 119 and that
said acoustic environment is largely absent of reverberation.
[0138] The masking module 124 adjusts the level of the
audio signal input so that it is only as loud as needed to mask
the residual noise. Generally, in the example above, if the
ambient noise level Was 55 dB rather than 60 dB SPL, then
the audio signal Would be presented to the user at a level of
45 dB rather than 50 dB SPL.
[0139] The masking module 124 adjusts a gain applied to
a signal multiplier 410 in a feedback arrangement based on
In such conditions continuous constant-level noise and for
the resulting microphone input 119. In general, the amount
SNR<<0 dB provides good correlation betWeen the input
audio envelope (adjusted by the aforementioned delay) and
of gain determined by the module is based on the psychoa
coustic principles that aim to relate the degree of intelligi
the SNSR so that an appropriate gain to achieve high
compression of the audio input can be determined from the
SNSR. Examples of environments in Which NAUC may be
bility of speech signals in the face of interfering signals such
advantageously applied include telephone receivers, auto
mobiles, aircraft cockpits, hearing aids, and small limited
as noise and reverberation. One objective predictor of such
intelligibility is the Speech Transmission Index, Which is an
estimate of intelligibility based on a degree to Which the
reverberation rooms.
modulations of energy in speech (i.e., the energy envelope)
is preserved betWeen a desired signal and the signal pre
3 Auto-Masking (FIG. 4)
sented to the user. Such an index can be computed separately
at di?ferent frequencies or across a Wide frequency band.
[0135] The masking module 124 automatically adjusts the
[0140] Referring to FIG. 4, the masking-module 124
audio level to reduce or eliminate distraction or other
interference to the user from signal the residual ambient
noise in the earpiece. Such distraction is most commonly
caused by the conversation of nearby people, though other
sounds can also distract the user, for example While the user
is performing a cognitive task.
[0136] One approach to reducing or eliminating the dis
traction is to adjust the audio level to be sufficiently loud to
completely mask the residual ambient noise at all times. The
masking module 124 achieves a reduction or elimination of
the distraction Without requiring as loud a level. Generally,
the masking module 124 automatically determines an audio
level to provide partial masking of the residual noise that is
su?icient to prevent the noise (e.g., conversation) from
intruding on the user’s attention. This approach to removing
determines energy envelopes associated With each of the
microphone input 119 and the audio signal 125 after the gain
adjustment (at multiplier 410). The masking module 124
determines the amount of gain to apply based on the
relationship betWeen these energy envelopes. The gain is
adjusted in a feedback arrangement to maintain a desired
relationship betWeen the energy envelopes.
[0141] The audio signal 125 and the microphone input 119
are passed to band-pass ?lters 412 and 416, respectively. The
pass bands of these ?lters are l kHz-3 kHZ, Which is a band
Within Which speech energy contributes signi?cantly to
intelligibility. The ?ltered audio signal and microphone
distraction can be e?fective if the user has selected audio to
input are passed to envelope detectors 414 and 418, respec
tively. The envelope detectors perform a short-time averag
ing of the signal energy (i.e., squared amplitude) over a time
constant of approximately 10 ms, Which captures speech
listen to Which is inherently less distracting and to the user’s
modulations at rates of up to approximately 15 HZ.
Nov. 23, 2006
US 2006/0262938 A1
[0142]
The outputs of the tWo envelope detectors 414 and
418 are input to a correlator 420, Which provides an output
based on a past block length, Which in this version of the
system is chosen to be of duration 200 ms. The correlator
normalizes the tWo inputs to have the same average level
over the block length then computes the sum of the product
an energy calculation by loW-pass ?ltering With 10 ms time
constant the square of the ?ltered signal level. Alternatively,
the absolute value of the ?lter output can be loW-pass ?ltered
to determine an envelope. Also, other loW-pass ?lter time
constants than 10 ms may be used.
of those recent normalized envelope values. In general, if the
[0148]
correlation is high, then the microphone input largely results
be used. Alternatively, the correlation may use a non
from the audio input, Which means there is relatively little
residual noise (distracting conversation) present. If the cor
Other correlation block lengths than 200 ms may
rectangular (Weighted) WindoW.
relation is loW, the microphone input largely results from the
[0149]
residual noise and the input audio is not loud enough to
obscure it.
[0143] The output of the correlator 420 is subtracted at an
the audio to maintain a target correlation value betWeen the
adder 422 from a correlation target value. This value is set
based on a value determined experimentally to provide
su?icient masking of distracting speech. A typical value for
the correlation target is 0.7. Optionally, the user can adjust
the correlation target value based on the user’s preference,
the speci?c nature of the ambient noise, etc.
[0144] The output of the adder 422 is passed to an inte
grator 424. The integrator responds to a constant difference
betWeen the measured correlation and the target With a
steadily increasing (or decreasing, depending on the sign of
the difference) gain command. The gain command output of
the integrator 424 is applied to a multiplier 410, Which
adjusts the gain of the audio signal input. The integrator time
The embodiment above adjusts the volume level of
band-limited signal envelopes of the audio input and moni
tored microphone signal. Alternatively; the auto-masking
system could be designed to adjust the volume level to
maintain a target SNSR or SNR value.
[0150] The embodiment described above implements the
auto-masking system for use With headphones. Altema
tively, auto-masking could be implemented in other situa
tions, for example in situations that are characteriZed by an
approximately knoWn time delay for propagation of output
audio signal 125, through an acoustic environment, to
microphone signal 119 and an acoustic environment that is
largely absent of reverberation. Under such conditions auto
masking could be made to operate advantageously in a small
constant is chosen to establish a subjectively preferred rate
room.
at Which the audio gain controlling feedback loop shoWn in
FIG. 4 responds to changes in distracting conversation level.
4 Noise reduction (FIG. 5)
A response time of ?ve to ten seconds is appropriate.
[0151]
Alternative responses may be used in place of integrator
424. For example, a loW-pass ?lter With high gain at DC may
be used to regulate the output of correlator 420 to be
audio signal 125, Which has already been subject to gain
su?iciently close to the target value as to achieve the desired
level of masking.
3.1 Alternatives
[0145] To prevent dynamics in music used as masking
audio from intruding too much into one’s attention (e.g.,
When it is desired for the music to remain a pleasant
background to cognitive tasks) it may be desirable to com
press input audio 123 prior to the level adjustment provided
by the masking system of FIG. 4. A standard compressor
structure With compression ratio of 2:1 to 3:1 can be
appropriate (rather than the NAUC system described ear
lier), though some users may prefer other ratios, the NAUC
system, or perhaps no compression. The choice of type of
compression used can be made user selectable.
[0146]
Variations on the approach shoWn in FIG. 4 are
possible. Left and right earpiece microphone and audio
signals can be acted on separately or combined and the
monaural component processed to determine the gain to
apply to the audio. Multiple BPF pass-bands could be set
and the envelope detection and correlation done in parallel
on the different bands, With the resulting correlation factors
combined in a Weighted fashion prior to comparison With a
target. If random or natural sounds are desired as the
masking signal rather than music, these could be stored in
some compressed form in the system so that auto-masking
can be accomplished Without the need to connect to an audio
source.
[0147]
The embodiment described above determines the
audio and microphone envelopes (time-varying levels) from
The noise reduction module 126 is applied to the
control and/or compression. Referring to FIG. 5, the noise
canceller makes use of a negative feedback arrangement in
Which the microphone input 119 is fed back and compared
to a desired audio signal, and the difference is fed forWard
to the audio driver. This arrangement is similar to that taught
in Us. Pat. No. 4,455,675, issued to Bose and Carter, Which
is incorporated herein by reference. In FIG. 5, the feedback
loop includes control rules 520, Which provide gain and
frequency-dependent transfer function to be applied to the
electrical signal. The output 127 of the control rules 520 is
applied to the driver 116 in the earpiece. The driver has a
frequency-dependent transfer function D betWeen its elec
trical input 127 and the sound pressure 525 achieved in the
earpiece. The microphone 118 senses the sound pressure and
produces the electrical microphone input 119. The micro
phone has a transfer function M betWeen the sound pressure
526 and the resulting electrical microphone signal 119. A
preemphasis component 518 receives the output 125 from
the masking module 124 and passes its output to the feed
back loop. The preemphasis component 518 compensates
for non-uniform frequency response characteristics intro
duced by the feedback loop.
[0152] Based on this arrangement, the audio signal applied
to the noise canceller has an overall transfer function of
Nov. 23, 2006
US 2006/0262938 A1
While the ambient noise has a transfer function
other applications besides headphones Where the approaches
can be applied are telephones (?xed or mobile), automobiles
or aircraft cockpits, hearing aids, and small rooms.
1
(1+CMD]
[0158]
It is to be understood that the foregoing description
is intended to illustrate and not to limit the scope of the
invention, Which is de?ned by the scope of the appended
claims. Other embodiments are Within the scope of the
thereby attenuating the ambient noise beyond that Which is
achieved by the physical characteristics of the earpiece.
folloWing claims.
5 Implementation
What is claimed is:
[0153] The approaches described above are implemented
using analog circuitry, digital circuitry or a combination of
the tWo. Digital circuitry can include a digital signal pro
cessor that implements one or more of the signal processing
steps described above. In the case of an implementation
using digital signal processing, additional steps of anti-alias
?ltering and digitiZation and digital-to-analog conversion
1. A method for processing an audio signal comprising:
receiving the audio signal;
monitoring an acoustic signal that includes components of
an interfering signal and the audio signal;
generating a processed audio signal including compress
ing the audio signal at a ?rst compression ratio When
are not shoWn in the diagrams or discussed above, but are
applied in a conventional manner. The analog circuitry can
the audio signal is at a ?rst level determined from the
include elements such as discrete components, integrated
circuits such as operational ampli?ers, or large-scale analog
signal at a second compression ratio When the audio
signal is above a second level determined from the
integrated circuits.
monitored acoustic signal, the ?rst level being loWer
than the second level and the ?rst compression ratio
being at least three times greater than the second
[0154] The signal processor can be integrated into the
headphone unit, or alternatively, all or part of the processing
described above is housed in separate units, or housed in
conjunction With the audio source. An audio source for noise
masking can be integrated into the headphone unit thereby
avoiding the need to provide an external audio source.
[0155]
In implementations that make use of program
mable processors, such as digital signal processors or gen
eral purpose microprocessor, the system includes a storage,
monitored acoustic signal and compressing the audio
compression ratio.
2. The method of claim 1 Wherein generating the pro
cessed audio signal further comprises selecting a compres
sion ratio according to a relationship betWeen a level of the
audio signal and a level of the acoustic signal.
3. The method of claim 2 further comprising determining
the relationship betWeen the level of the audio signal and the
level of the acoustic signal Without separating the compo
such as a non-volatile semiconductor memory (e.g., “?ash”
memory) that holds instructions that When executed on the
nents of the interfering signal and the audio signal.
processor implement one or more of the modules of the
cessed audio signal comprises reducing a masking e?fect
related to the interfering signal.
system. In implementations in Which an audio source is
integrated With the headphone unit, such storage may also
hold a digitiZed version of the audio signal input, or may
hold instructions for synthesizing such an audio signal.
6 Alternatives
[0156] The discussion above concentrates on processing
of a single channel. For stereo processing (i.e., tWo channels,
4. The method 6 f claim 1 Wherein generating the pro
5. The method of claim 4 Wherein reducing the masking
e?fect related to the interfering signal comprises at least one
of reducing an intelligibility of the interfering signal, reduc
ing a distraction by the interfering signal, and partially
masking the interfering signal.
separate instance of signal processors for each ear/channel.
6. The method of claim 1 Wherein generating the pro
cessed audio signal comprises adjusting at least one of a gain
and a compression of the audio signal according to a
masking effect related to the interfering signal and to the
Alternatively, some or all of the processing is shared for the
audio signal.
one associated With each ear), one approach is to use a
tWo channels. For example, the audio inputs and microphone
inputs may be summed for the tWo channels and a common
gain is then applied to both the right and the left audio
inputs. Some of the processing steps may be shared betWeen
the channels While others are done separately. In the present
embodiment the compression and masking stages are per
formed on a monaural channel While the active noise reduc
7. The method of claim 1 Wherein the second compression
ratio is approximately one to one.
8. The method of claim 1 Wherein the second compression
ratio is less than tWo to one.
9. The method of claim 1 Wherein the ?rst compression
ratio is at least three to one.
10. The method of claim 1 Wherein the ?rst compression
tion is performed separately for each channel.
ratio is at least ?ve to one.
[0157] Although aspects of the system, including both
upWard compression (NAUC) and auto-masking, are
signal further comprises applying the second compression
described above in the context of driving headphones, the
approaches can be applied in other environments. Prefer
ably, such other environments are ones in Which (a) the
microphone can sense What is being heard at the ear of users,
(b) time delays in propagation of audio from speakers to the
microphone are small compared to envelope detector time
constants and (c) there is little reverberation. Examples of
11. The method of claim 1 Wherein compressing the audio
ratio When a level of the audio signal is at least 10 dB above
a level of the interfering signal.
12. The method of claim 1 further comprising transmitting
the processed audio signal to an earpiece.
13. The method of claim 12 Wherein monitoring the
acoustic signal comprises monitoring the acoustic signal in
the earpiece.
Nov. 23, 2006
US 2006/0262938 A1
14. The method of claim 12 wherein a source of the
interfering signal is outside of the earpiece.
15. The method of claim 1 Wherein the acoustic signal
includes at least some component of the audio signal.
16. The method of claim 15 Wherein monitoring the
acoustic signal comprises monitoring the acoustic signal
outside an earpiece.
17. The method of claim 1 further comprising applying
active noise reduction according to the acoustic signal.
18. The method of claim 1 further comprising determining
a time-varying relationship betWeen a level of the audio
signal and a level of the acoustic signal.
19. The method of claim 18 Wherein generating the
processed audio signal comprises varying a gain of the audio
signal over time according to the time-varying relationship.
20. The method of claim 18 Wherein generating the
processed audio signal comprises varying a degree of com
pression of the audio signal over time according to the
time-varying relationship.
21. The method of claim 1 Wherein generating the pro
cessed audio signal further comprises expanding the audio
signal When the audio signal is beloW a threshold level.
30. The audio processing system of claim 22 further
comprising:
a masking module that receives the audio signal and the
acoustic signal, the masking module including circuitry
for processing the audio signal according to a level of
the acoustic signal, including controlling a level of the
audio signal input to reduce a masking effect of an
interfering signal present in the acoustic signal.
31. The audio processing system of claim 30 further
comprising a selector to selectively enable at least one of the
compression circuit and the masking module.
32. A method for audio processing comprising:
receiving an audio signal;
monitoring an acoustic signal that is related to the audio
signal;
determining a threshold level according to a relationship
betWeen a level of the audio signal and a level of the
acoustic signal; and
processing the audio signal by compressing the audio
signal When the threshold level is beloW a ?rst level and
22. An audio processing system comprising:
maintaining the audio signal substantially unmodi?ed
an input for receiving an audio signal;
When the threshold level is above a second level.
a microphone for monitoring an acoustic signal, the
acoustic signal including components of an interfering
signal and the audio signal;
a compressor circuit for compressing the audio signal at
a ?rst compression ratio When the audio signal is at a
?rst level determined from the monitored acoustic
signal and compressing the audio signal at a second
compression ratio When the audio signal is above a
second level determined from the monitored acoustic
signal, the ?rst level being loWer than the second level
and the ?rst compression ratio being at least three times
greater than the second compression ratio.
23. The audio processing system of claim 22 Wherein the
compressor circuit is con?gured to reduce a masking e?fect
related to the interfering signal.
24. The audio processing system of claim 23 Wherein
reducing the masking e?fect related to the interfering signal
33. The method of claim 32 Wherein processing the audio
signal further comprises reducing a masking effect of the
interfering signal in response to the threshold level.
34. The method of claim 33 Wherein reducing the masking
e?fect comprises at least one of reducing an intelligibility of
the interfering signal, reducing a distraction by the interfer
ing signal, and partially masking the interfering signal.
35. The method of claim 33 Wherein determining a
threshold level comprises determining a relationship
betWeen a level of the audio signal and a level of the acoustic
signal Without separating the components related to the
audio signal and an interfering signal.
36. The method of claim 32 Wherein determining a
threshold level comprises determining according to a rela
tionship betWeen a level of the audio signal and a level of the
acoustic signal Without separating the components related to
the audio signal and an interfering signal.
interfering signal, reducing a distraction by the interfering
37. The method of claim 32 Wherein compressing the
audio signal When the threshold level is beloW a ?rst level
comprises applying a compression ratio that is at least three
signal, and partially masking the interfering signal.
to one.
25. The audio processing system of claim 23 further
comprising a tracking circuit con?gured to determine a
relationship betWeen a level of the audio signal and a level
of the acoustic signal Without separating the components of
38. The method of claim 32 Wherein compressing the
audio signal When the threshold level is beloW a ?rst level
comprises applying a compression ratio that is at least ?ve
comprises at least one of reducing an intelligibility of the
the audio signal and the interfering signal.
26. The audio processing system of claim 22 Wherein the
second level is greater than the ?rst level.
27. The audio processing system of claim 22 Wherein the
acoustic signal monitored by the microphone includes a at
least some component of the audio signal.
28. The audio processing system of claim 22 further
comprising an earpiece containing the microphone and a
driver.
29. The audio processing system of claim 22 Wherein at
least one of the tracking circuit and the compressor circuit is
at least partially contained Within the earpiece.
to one.
39. The method of claim 32 Wherein maintaining the
audio signal substantially unmodi?ed comprises passing the
audio signal Without substantial compression.
40. The method of claim 39 Wherein passing the audio
signal Without substantial compression comprises applying a
compression ratio that is approximately one to one.
41. The method of claim 32 Wherein the threshold level
corresponds to the second level When a level of the audio
signal is at least 10 dB above a level of an interfering signal.
42. The method of claim 32 further comprising determin
ing a level of an interfering signal based on a level of the
acoustic signal and a level of the audio signal.
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