 |
Claims  |
|
|
What is claimed is:
1. An echo cancelling device for reducing the effects of acoustic feedback
between a loudspeaker and microphone in a communication system,
comprising:
a first signal splitter for separating a microphone signal into a plurality
of bandlimited microphone signals,
a second signal splitter for separating a loudspeaker signal into a
plurality of bandlimited loudspeaker signals, the band of frequencies of
each bandlimited loudspeaker signal being approximately the same as the
band of frequencies of a corresponding bandlimited microphone signal,
a plurality of band echo estimators, each band echo estimator for
generating an echo estimation signal for a bandlimited loudspeaker signal,
said echo estimation signal representing an approximation of the acoustic
feedback of said bandlimited loudspeaker signal into a corresponding
bandlimited microphone signal,
at least one subtractor for subtracting an echo estimation signal from a
bandlimited microphone signal of the same frequency band as the echo
estimation signal to produce a bandlimited echo corrected microphone
signal,
means for estimating whether a first bandlimited echo corrected microphone
signal is substantially derived from acoustic feedback between said
loudspeaker and said microphone, and
at least one signal clipper for attenuating said first bandlimited echo
corrected microphone signal during periods of time during which said first
bandlimited echo corrected microphone signal is substantially derived from
acoustic feedback between said loudspeaker and microphone.
2. The echo cancelling device of claim 1 wherein each said signal clipper
comprises a gain adjustment module for gradually increasing the
attenuation of said signal clipper following the commencement of each said
period during which said bandlimited echo corrected signal is
substantially derived from acoustic feedback.
3. The echo cancelling device of claim 2 wherein each said gain adjustment
module gradually decreases the attenuation of said signal clipper
following the termination of each said period during which said
bandlimited echo corrected signal is substantially derived from acoustic
feedback.
4. The echo cancelling device of claim 1 further comprising a plurality of
noise fillers, each noise filler comprising:
means for generating a background noise estimation signal representing the
background noise contained in a bandlimited microphone signal, and
means for adding said background estimation signal to said bandlimited
microphone signal to compensate for attenuation of said background noise
by said at least one signal clipper.
5. The echo cancelling device of claim 4 wherein each noise filler adds
said background estimation signal to said attenuated bandlimited echo
corrected microphone signal in an amount complementary to the magnitude of
said attenuation.
6. The echo cancelling device of claim 4 wherein each said means for
generating a background noise estimation signal comprises:
means for estimating the energy of a background noise component of said
bandlimited microphone signal, and
means for generating said background noise estimation signal having an
energy approximately equal to said estimated energy of said background
noise component.
7. The echo cancelling device of claim 6 wherein said means for estimating
the energy of a background noise component of said bandlimited microphone
signal comprises:
means for estimating the interval energy of said bandlimited microphone
signal in each of a plurality of time intervals, and
means for selecting the minimum estimated energy from said estimated
interval energies for use as said estimated energy of said background
noise component.
8. An echo cancelling device for reducing the effects of acoustic feedback
between loudspeaker and microphone in a communication system, comprising:
a first signal splitter for separating a near end microphone signal into a
plurality of bandlimited microphone signals,
a second signal splitter for separating a loudspeaker signal into a
plurality of bandlimited loudspeaker signals, the band of frequencies of
each bandlimited loudspeaker signal being approximately the same as the
band of frequencies of a corresponding bandlimited microphone signal,
a plurality of adaptive echo estimators, each adaptive echo estimator for
generating an echo estimation signal for an associated bandlimited
loudspeaker signal, said echo estimation signal representing an
approximation of the acoustic feedback of said bandlimited loudspeaker
signal into a corresponding bandlimited microphone signal,
at least one subtractor for subtracting an echo estimation signal from an
associated bandlimited microphone signal having the same frequency band as
said echo estimation signal, to produce a bandlimited echo corrected
microphone signal,
at least one local speech detector for identifying periods of time during
which said near end microphone signal is substantially derived from
acoustic feedback between said loudspeaker and microphone, and
at least one adjustment module for adjusting characteristics of at least
one said adaptive echo estimator during said identified periods of time.
9. The echo cancelling device of claim 8 wherein said at least one local
speech detector comprises:
means for computing an attenuated version of a bandlimited loudspeaker
signal, and
means for comparing said attenuated bandlimited loudspeaker signal to a
bandlimited echo corrected microphone signal to determine whether said
bandlimited echo corrected microphone signal is substantially derived from
acoustic feedback between said loudspeaker and microphone.
10. The echo cancelling device of claim 9 wherein said means for computing
an attenuated version of said bandlimited loud speaker signal comprises:
means for calculating a dynamic gain which varies with the magnitude of
past samples of said bandlimited loudspeaker signal, and
means for attenuating said bandlimited loudspeaker signal based on the
value of said dynamic gain.
11. The echo cancelling device of claim 10 wherein said means for
calculating a dynamic gain comprises:
means for determining the magnitude of the most recent peak of said
bandlimited loudspeaker signal, and
means for attenuating said dynamic gain based on the length of time since
said most recent peak occurred in said bandlimited loudspeaker signal.
12. The echo cancellation device of claim 10 wherein said means for
calculating a dynamic gain comprises:
means for setting said dynamic gain equal to the current value of said
bandlimited loudspeaker signal if said current value of said bandlimited
loudspeaker signal is greater than said dynamic gain.
13. The echo cancellation device of claim 12 wherein said means for
calculating a dynamic gain further comprises means for attenuating said
dynamic gain at a rate approximately equal to the rate at which
reverberance of said bandlimited loudspeaker signal dampens.
14. The echo cancelling device of claim 9 wherein said means for computing
an attenuated version of said bandlimited loudspeaker signal comprises:
means for estimating the effective gain of the acoustic channel between
said loudspeaker and microphone, and
means for attenuating said bandlimited loudspeaker signal based o the value
of said effective gain.
15. The echo cancelling device of claim 8 wherein said at least one local
speech detector comprises a plurality of local speech detectors, each for
identifying periods of time during which an associated bandlimited echo
corrected microphone signal is substantially derived from acoustic
feedback between said loudspeaker and microphone.
16. The echo cancelling device of claim 8 and further comprising at least
one divergence detector for determining when an adaptive echo estimator is
diverging, said at least one adjustment module further comprising a means
for resetting the characteristics of an echo estimator which is diverging.
17. The echo cancelling device of claim 16 wherein said at least one
divergence detector comprises means for comparing the energy of a
bandlimited microphone signal to the energy of a corresponding bandlimited
echo corrected microphone signal, and means for declaring when said
adaptive echo estimator is diverging based on said comparison.
18. An echo cancelling device for reducing the effects of acoustic feedback
between a loudspeaker and microphone in a communication system,
comprising:
a first whitening filter for receiving a microphone signal and generating
in response thereto a whitened version of said microphone signal,
a first signal splitter for separating said whitened microphone signal into
a plurality of bandlimited microphone signals,
a second whitening filter for receiving a loudspeaker signal and generating
in response thereto a whitened version of said loudspeaker signal,
a second signal splitter for separating said whitened loudspeaker signal
into a plurality of bandlimited loudspeaker signals, the band of
frequencies of each bandlimited loudspeaker signal being approximately the
same as the band of frequencies of a corresponding bandlimited microphone
signal,
a plurality of adaptive band echo estimators, each adaptive band echo
estimator for generating an echo estimation signal for a bandlimited
loudspeaker signal, said echo estimation signal representing an
approximation of the acoustic feedback of said bandlimited loudspeaker
signal into a corresponding bandlimited microphone signal,
at least one subtractor for subtracting an estimation signal from a
bandlimited microphone signal of the same frequency band as the estimation
signal to produce a bandlimited echo corrected microphone signal.
19. The echo cancelling system of claim 18 further comprising:
signal composer for composing said bandlimited echo correction microphone
signals into a composite microphone signal, and
an inverse whitening filter for receiving said composite microphone signal
and performing the inverse operation of said first whitening filter.
20. An echo cancelling device for reducing the effects of acoustic feedback
between a loudspeaker and microphone in a communication system,
comprising:
a simple whitening filter having a transfer function with approximately one
zero, said whitening filter for receiving a microphone signal and
generating in response thereto a whitened version of said microphone
signal,
at least one adaptive filter for generating an echo estimation signal
representing an approximation of a component of said whitened version of
said microphone signal due to acoustic feedback between said loudspeaker
and microphone,
a subtractor for removing said echo estimation signal from said whitened
version of said microphone signal to yield an echo corrected microphone
signal.
21. The echo cancelling device of claim 20 wherein said simple whitening
filter has a transfer function h(z)=a-b/z where a is approximately 1 and b
is approximately 0.95.
22. The echo cancelling device of claim 21 further comprising:
an inverse whitening filter, having a transfer function g(z)=c/(d-e/z)
where c and d are approximately equal to one and e is approximately equal
to 0.95, for receiving said echo corrected microphone signal and
generating in response thereto at unwhitened version thereof.
23. An echo cancelling device for reducing the effects of acoustic feedback
between a loudspeaker and microphone in a communication system,
comprising:
a first whitening filter for receiving a near end microphone signal and
generating in response thereto a whitened version of said near end
microphone signal,
a first signal splitter for separating said whitened microphone signal into
a plurality of bandlimited microphone signals,
a second whitening filter for receiving a loudspeaker signal and generating
in response thereto a whitened version of said loudspeaker signal,
a second signal splitter for separating said whitened loudspeaker signal
into a plurality of bandlimited loudspeaker signals, the band of
frequencies of each bandlimited loudspeaker signal being approximately the
same as the band of frequencies of a corresponding bandlimited microphone
signal,
a plurality of adaptive echo estimators, each adaptive echo estimator for
generating an echo estimation signal for an associated bandlimited
loudspeaker signal, said echo estimation signal representing an
approximation of the acoustic feedback of said bandlimited loudspeaker
signal into a corresponding bandlimited microphone signal,
a subtractor means for subtracting from each bandlimited microphone signal
an echo estimation signal having the same frequency band as the
bandlimited microphone signal, to produce a bandlimited echo corrected
microphone signal,
a plurality of local speech detectors each for identifying periods of time
during which a corresponding bandlimited microphone signal is
substantially derived from acoustic feedback between said loudspeaker and
microphone, and
a plurality of adjustment modules each for adjusting characteristics of at
least one said adaptive echo estimator during said identified periods of
time when a corresponding bandlimited microphone signal is substantially
derived from acoustic feedback, and
a plurality of signal clippers, each clipper for attenuating a
corresponding bandlimited echo corrected microphone signal during periods
of time during which said bandlimited echo corrected microphone signal is
substantially derived from acoustic feedback between said loudspeaker and
microphone, and
a plurality of noise fillers, each noise filler comprising:
a means for generating a background noise estimation signal representing
the background noise contained in a bandlimited microphone signal, and
a means for adding said background estimation signal to said attenuated
bandlimited echo corrected microphone signal in an amount complementary to
the magnitude of said attenuation.
24. A method for reducing the effects of acoustic feedback between a
loudspeaker and microphone in a communication system, comprising the steps
of:
separating a microphone signal into a plurality of bandlimited microphone
signals,
separating a loudspeaker signal into a plurality of bandlimited loudspeaker
signals, the band of frequencies of each bandlimited loudspeaker signal
being approximately the same as the band of frequencies of corresponding
bandlimited microphone signal,
generating an echo estimation signal for each bandlimited loudspeaker
signal, each said echo estimation signal representing an approximation of
the acoustic feedback of a bandlimited loudspeaker signal into a
corresponding bandlimited microphone signal,
subtracting an echo estimation signal from a bandlimited microphone signal
of the same frequency band as the echo estimation signal to produce a
bandlimited echo corrected microphone signal, and
attenuating said bandlimited echo corrected microphone signal during
periods of time during which said bandlimited echo corrected microphone
signal is substantially derived from acoustic feedback between said
loudspeaker and microphone.
25. The method of claim 24 wherein said attention step comprises gradually
increasing the attenuation of said bandlimited echo corrected signal
following the commencement of each said period during which said
bandlimited echo corrected signal is substantially derived from acoustic
feedback.
26. The method claim 25 wherein said attenuation step further comprises
gradually decreasing the attenuation of said bandlimited echo corrected
signal following the termination of each said period during which said
bandlimited echo corrected signal is substantially derived from acoustic
feedback.
27. The method claim 24 further comprising the steps of:
generating a background noise estimation signal representing the background
noise contained in a bandlimited microphone signal, and
adding said background estimation signal to said attenuated bandlimited
echo corrected microphone signal during said period of time during which
said bandlimited echo corrected signal is substantially derived from
acoustic feedback.
28. The method of claim 27 wherein said background estimation signal is
added to said attenuated bandlimited echo corrected microphone signal in
an amount complementary to the magnitude of said attenuation.
29. The method of claim 28 wherein generating a background noise estimation
signal comprises the steps of:
estimating the energy of a background noise component of said bandlimited
microphone signal, and
generating said background noise estimation signal having an energy
approximately equal to said estimated energy of said background noise
component.
30. The method of claim 29 wherein estimating the energy of a background
noise component of said bandlimited microphone signal comprises the steps
of:
estimating the interval energy of said bandlimited microphone signal in
each of a plurality of time intervals,
selecting the minimum estimated energy from said estimated interval
energies for use as said estimated energy of said background noise
component.
31. A method for reducing the effects of acoustic feedback between
loudspeaker and microphone in a communication system, comprising the steps
of:
separating a near end microphone signal into a plurality of bandlimited
microphone signals,
separating a loudspeaker signal into a plurality of bandlimited loudspeaker
signals, the band of frequencies of each bandlimited loudspeaker signal
being approximately the same as the band of frequencies of a corresponding
bandlimited microphone signal,
applying each echo bandlimited loudspeaker signal to a corresponding
adaptive echo estimator which in response generates an echo estimation
signal, each said echo estimation signal representing an approximation of
the acoustic feedback of said bandlimited loudspeaker signal into a
corresponding bandlimited microphone signal,
subtracting an echo estimation signal from an associated bandlimited
microphone signal having the same frequency band as said echo estimation
signal, to produce a bandlimited echo corrected microphone signal,
identifying periods of time during which said near end microphone signal is
substantially derived from acoustic feedback between said loudspeaker and
microphone, and
adjusting characteristics of at least one said adaptive echo estimator
during said identified periods of time.
32. The method of claim 31 wherein identifying periods of time comprises
the steps of:
computing an attenuated version of a bandlimited loudspeaker signal, and
comparing said attenuated bandlimited loudspeaker signal to a bandlimited
echo corrected microphone signal to determine whether said bandlimited
echo corrected microphone signal is substantially derived from acoustic
feedback between said loudspeaker and microphone.
33. The method of claim 32 wherein computing an attenuated version of said
bandlimited loudspeaker signal comprises the steps of:
calculating a dynamic gain which varies with the magnitude of past samples
of said bandlimited loudspeaker signal, and
attenuating said bandlimited loudspeaker signal based on the value of said
dynamic gain.
34. The echo cancelling device of claim 33 wherein calculating a dynamic
gain comprises the steps of:
determining the magnitude of the most recent peak of said bandlimited
loudspeaker signal, and
attenuating said dynamic gain based on the length of time since said most
recent peak occurred in said loudspeaker signal.
35. The method of claim 33 wherein calculating a dynamic gain comprises the
steps of:
setting said dynamic gain equal to the current value of said bandlimited
loudspeaker signal if said current value of said bandlimited loudspeaker
signal is greater than said dynamic gain.
36. The method of claim 35 wherein calculating a dynamic gain further
comprises attenuating said dynamic gain at a rate approximately equal to
the rate at which reverberance of said bandlimited loudspeaker signal
dampens.
37. The method of claim 32 wherein computing an attenuated version of said
bandlimited loudspeaker signal comprises the steps of:
estimating the effective gain of the acoustic channel between said
loudspeaker and microphone, and
attenuating said bandlimited loudspeaker signal based on the value of said
effective gain.
38. A method for reducing the effects of acoustic feedback between a
loudspeaker and microphone in a communication system, comprising the steps
of:
receiving a microphone signal and generating in response thereto a whitened
version of said microphone signal,
separating said whitened microphone signal into a plurality of bandlimited
microphone signals,
receiving a loudspeaker signal and generating in response thereto a
whitened version of said loudspeaker signal,
separating said whitened loudspeaker signal into a plurality of bandlimited
loudspeaker signals, the band of frequencies of each bandlimited
loudspeaker signal being approximately the same as the band of frequencies
of a corresponding bandlimited microphone signal,
generating an echo estimation signal for each bandlimited loudspeaker
signal, each said echo estimation signal representing an approximation of
the acoustic feedback of said bandlimited loudspeaker signal into a
corresponding bandlimited microphone signal,
subtracting each estimation signal from a bandlimited microphone signal of
the same frequency band as the estimation signal to produce a bandlimited
echo corrected microphone signal.
39. The method of claim 38 further comprising the steps of:
composing said bandlimited echo correction microphone signals into a
composite microphone signal, and
performing the inverse operation of said first whitening filter on said
composite microphone signal.
40. An method for reducing the effects of acoustic feedback between a
loudspeaker and microphone in a communication system, comprising the steps
of:
receiving a near end microphone signal and generating in response thereto a
whitened version of said near end microphone signal,
separating said whitened microphone signal into a plurality of bandlimited
microphone signals,
receiving a loudspeaker signal and generating in response thereto a
whitened version of said loudspeaker signal,
separating said whitened loudspeaker signal into a plurality of bandlimited
loudspeaker signals, the band of frequencies of each bandlimited
loudspeaker signal being approximately the same as the band of frequencies
of a corresponding bandlimited microphone signal,
generating an echo estimation signal for each bandlimited loudspeaker
signal, each said echo estimation signal representing an approximation of
the acoustic feedback of said bandlimited loudspeaker signal into a
corresponding bandlimited microphone signal,
subtracting each estimation signal from an associated bandlimited
microphone signal having the same frequency band as the estimation signal,
to produce a bandlimited echo corrected microphone signal,
identifying periods of time during which a corresponding bandlimited
microphone signal is substantially derived from acoustic feedback between
said loudspeaker and microphone,
attenuating a corresponding bandlimited echo corrected microphone signal
during periods of time during which said bandlimited echo corrected
microphone signal is substantially derived from acoustic feedback between
said loudspeaker and microphone,
generating a background noise estimation signal representing the background
noise contained in a bandlimited microphone signal, and
adding said background estimation signal to said attenuated bandlimited
echo corrected microphone signal in an amount complementary to the
magnitude of said attenuation. |
|
 |
|
 |
Claims |
|
|
|
Description |
|
|
This specification includes a microfiche appendix having three sheets of
microfiche which collectively contain two hundred and thirteen frames.
BACKGROUND OF THE INVENTION
The invention relates generally to reducing unwanted audio or acoustic
feedback in a communication system, and particularly to an adaptive
acoustic echo cancellation device for suppressing acoustic feedback
between the loudspeaker and microphone of a telephone unit in a
teleconferencing system. The telephone unit of a typical audio
conferencing system includes a loudspeaker for broadcasting an incoming
telephone signal into an entire room. Similarly, the telephone's
microphone is typically designed to pick up the voice of any person within
the room and transmit the voice to a remote telephone at the far end of
the communication system.
Unlike conventional hand held telephone sets, conference telephone units
are prone to acoustic feedback between the loudspeaker unit and
microphone. For example, a voice signal which is broadcast into the room
by the loudspeaker unit may be picked up by the microphone and transmitted
back over the telephone lines. As a result, persons at the far end of the
communication system hear an echo of their voice. The echo lags the
person's voice by the round trip delay time for the voice signal.
Typically, the echo is more noticeable as the lag between the person's
voice and the echo increases. Accordingly, it is particularly annoying in
video conferencing systems which transmit both video and audio information
over the same telephone lines. The additional time required to transmit
video data increases the round trip delay of the audio signal, thereby
extending the lag between a person's voice and the echo.
Many conference telephones avoid echo by allowing only half duplex
communication (that is, by allowing communication over the phone line to
occur in only one direction at a time) thereby preventing feedback. For
example, when the loudspeaker unit is broadcasting a voice, the telephone
disables the microphone to prevent the loudspeaker signal from being fed
back by the microphone.
While a half duplex system avoids echo, it often cuts off a person's voice
in mid-sentence. For example, when both parties speak simultaneously, the
telephone unit allows communication in only one direction, thereby
clipping the voice of one party.
Some loudspeaker telephones employ echo cancellation in an attempt to allow
full-duplex communication without echo. Conventional echo cancellation
devices attempt to remove from the microphone signal the component
believed to represent the acoustic feedback. More specifically, they
prepare an electric signal which duplicates the acoustic feedback between
the loudspeaker and the microphone. This electric signal is subtracted
from the microphone signal in an attempt to remove the echo.
Electrically duplicating the acoustic feedback is difficult since the
acoustic response of the room containing the microphone and speaker must
in essence be simulated electrically. This is complicated by variations in
the acoustic characteristics of different rooms and by the dramatic
changes in a given room's characteristics which occur if the microphone or
loudspeaker is moved, or if objects are moved in the room.
To compensate for the changing characteristics of the room, many echo
cancellation devices model the room's characteristics with an adaptive
filter which adjusts with changes in the room. More specifically, the
electric signal used to drive the telephone's loudspeaker is applied to a
stochastic gradient least-means-squares adaptive filter whose tap weights
are set to estimate the room's acoustic response. The output of the
filter, believed to estimate the acoustic echo, is then subtracted from
the microphone signal to eliminate the component of the microphone signal
derived from acoustic feedback. The resultant "echo corrected" signal is
then sent to listeners at the far end of the communication system.
To assure that the adaptive filter accurately estimates the room's
response, the device monitors the echo corrected signal. During moments
when no one is speaking into the microphone, the adaptive filter adjusts
its tap weights such that the energy of the echo corrected signal is at a
minimum. In theory, the energy of the echo corrected signal is minimized
when the adaptive filter removes from the microphone signal an accurate
replica of the acoustic feedback. However, the adaptive process must be
disabled whenever a person speaks into the microphone. Otherwise, the unit
will attempt to adjust the tap weights in an effort to eliminate the
speech.
Since a speech signal is highly correlated, the adaptive filter tends to
converge very slowly. Accordingly, some commercial echo cancellation
devices attempt to measure the room's acoustic response using a white
noise training sequence. During the training sequence, an unpleasant white
noise is emitted from the loudspeaker and is acoustically fed back to the
microphone. The white noise received by the microphone is a highly
uncorrelated signal, causing the adaptive filter to converge quickly. If
the filter loses convergence during the conversation, the training
sequence must be repeated, briefly interrupting conversation with an
annoying white noise signal.
Therefore, one object of the present invention is to provide an acoustic
echo cancellation device which allows full duplex communication while
reducing or eliminating echo. A further object is to eliminate the need
for a training sequence with a relative simple filter design which
converges quickly.
SUMMARY OF THE INVENTION
The invention relates to a method and apparatus for reducing acoustic
feedback in a full duplex communication system. The method includes
separating a near end microphone signal into a plurality of bandlimited
microphone signals, and similarly separating a near end loudspeaker signal
into a plurality of bandlimited loudspeaker signals. Each bandlimited
loudspeaker signal is filtered to generate an echo estimation signal which
represents an approximation of the acoustic feedback of the bandlimited
loudspeaker signal into the near end microphone signal. Each echo
cancellation signal is subtracted from the bandlimited microphone signal
whose frequency band includes the frequencies of the echo cancellation
signal, thereby removing an estimation of the echo in that frequency band.
In one embodiment, a plurality of adaptive filters, each having tap weights
which adapt with changes in the acoustic characteristics of the channel
between a loudspeaker and microphone are used to generate the echo
estimation signals. The performance of the adaptive filter for each band
is monitored to determine when the filter's tap weights are diverging. If
a gi | | |
