 |
Claims  |
|
|
What is claimed is:
1. A method for removing, from a communication signal, echoes generated
from a reference signal by feedback through a medium, comprising:
a) declaring one of a plurality of operating states, depending on data
characteristics of the communication signal and the reference signal;
b) responsive to the declared operating state, operating a main adaptive
filter such that data are acquired describing the transfer characteristics
of the medium in at least one signal frequency band;
c) responsive to said data, processing the reference signal to generate an
estimated echo signal;
d) subtracting said estimated echo signal from the communication signal,
thereby to form an echo-corrected signal;
e) delaying the communication signal and the reference signal such that
each said signal is used as input for step (a) before it is used as input
for step (b), such that the declaration of the operating state is made
before the generation of the estimated echo signal, wherein:
f) step (a) comprises detecting a condition, to be referred to as
"doubletalk," wherein the communication signal contains more activity than
can be accounted for solely by the echoes;
g) step (a) further comprises, when doubletalk is detected, the step of
declaring a corresponding operating state, to be referred to as the DT
state;
h) during at least the DT state, the main adaptive filter is not permitted
to acquire new data describing said transfer characteristics; and
i) during at least some states which include the DT state, bidirectional
communication is permitted.
2. A method for removing, from a communication signal, echoes generated
from a reference signal by feedback through a medium, comprising:
a) declaring one of a plurality of operating states, depending on data
characteristics of the communication signal and the reference signal;
b) responsive to the declared operating state, operating a main adaptive
filter such that data are acquired describing the transfer characteristics
of the medium in at least one signal frequency band;
c) responsive to said data, processing the reference signal to generate an
estimated echo signal; and
d) subtracting said estimated echo signal from the communication signal,
thereby to form an echo-corrected signal; wherein:
e) step (a) comprises detecting a condition, to be referred to as
"doubletalk," wherein the communication signal contains more activity than
can be accounted for solely by the echoes;
f) step (a) further comprises, when doubletalk is detected, the step of
declaring a corresponding operating state, to be referred to as the DT
state;
g) during at least the DT state, the main adaptive filter is not permitted
to acquire new data describing said transfer characteristics;
h) during at least some states which include the DT state, bidirectional
communication is permitted;
i) the doubletalk-detecting step comprises: operating an auxiliary adaptive
filter that is independent of said main adaptive filter and is permitted
to acquire new transfer-characteristic data in all of said operating
states; and evaluating the performance of the auxiliary adaptive filter;
and
j) the auxiliary adaptive filter acts upon a frequency subband smaller
than, and contained within, the signal frequency band.
3. The method of claims 1, or 2, wherein the doubletalk-detecting step
comprises the further steps of:
evaluating a statistic, to be denoted G, that indicates whether the
communication signal contains more energy than the expected echo generated
from the reference signal;
evaluating a statistic, to be denoted E, that indicates whether an echo
signal is being effectively removed during the echo-signal-subtracting
step; and
declaring the DT state only if both G and E satisfy predetermined
conditions.
4. The method of claim 3, wherein E and G are evaluated on respective
signals derived from the communication signal and the reference signal,
within a group of one or more frequency sub-bands that are each smaller
than, and contained within, the signal frequency band.
5. The method of claim 4, wherein said group of one or more frequency
subbands collectively spans a frequency range that is smaller than, and
contained within, the signal frequency band.
6. The method of claim 4, wherein the steps of evaluating E and G comprise
adaptively filtering the sub-band signal or signals derived from the
communication signal.
7. The method of claim 6, wherein, if the reference signal is denoted F,
the communication signal is denoted S, and the echo-corrected signal is
denoted T, then the step of evaluating G comprises the further steps of:
a) evaluating the statistic L'.sub.a, defined as
##EQU12##
wherein the maxima are taken over appropriate time windows; b) evaluating
the statistic L.sub.a, defined
##EQU13##
wherein the maxima are taken over appropriate time windows; c) evaluating
the statistic L.sub.a, defined as a gated average of L.sub.a that counts
only those values of L.sub.a that occur when E exceeds a predetermined
threshold;
d) setting G=1 if L'.sub.a >T.sub.G L.sub.a, wherein T.sub.G is a
predetermined threshold; and
e) setting G=0 otherwise.
8. The method of claim 7, wherein the reference signal comprises speech
received by a communication device from a remote user, the communication
signal comprises speech injected into the communication device by a local
user, the communication device includes a microphone and a loudspeaker,
and the step of operating a main adaptive filter is carried out in such a
way that data are acquired describing at least the transfer
characteristics of an acoustic path from the loudspeaker to the microphone
through a room that contains the communication device.
9. The method of claim 3, wherein the reference signal comprises speech
received by a communication device from a remote user, the communication
signal comprises speech injected into the communication device by a local
user, the communication device includes a microphone and a loudspeaker,
and the step of operating a main adaptive filter is carried out in such a
way that data are acquired describing at least the transfer
characteristics of an acoustic path from the loudspeaker to the microphone
through a room that contains the communication device.
10. The method of claim 1, or 2, wherein the reference signal comprises
speech received by a communication device from a remote user, the
communication signal comprises speech injected into the communication
device by a local user, the communication device includes a microphone and
a loudspeaker, and the step of operating a main adaptive filter is carried
out in such a way that data are acquired describing at least the transfer
characteristics of an acoustic path from the loudspeaker to the microphone
through a room that contains the communication device.
11. In the operation of a communication system having a microphone and a
loudspeaker so disposed that there is a feedback path from the loudspeaker
to the microphone, a method for measuring coupling loss that inheres
between the loudspeaker and the microphone, said loss optionally including
gain due to amplification of input to the loudspeaker and output from the
microphone, the method comprising:
a) operating an adaptive filter, thereby to acquire data that represent
coupling characteristics of the feedback path;
b) detecting when the adaptive filter is well adapted and when it is not
well adapted; and
c) updating an estimate of the coupling loss only when the adaptive filter
is well adapted, said updating step comprising the further step of
enabling a gated average to be updated only when an indicator of adaption
of said adaptive filter exceeds a threshold.
12. A method for operating an echo canceller for removing, from a
communication signal, echoes generated from a reference signal by feedback
through a medium, comprising:
a) declaring one of a plurality of operating states of the echo canceller,
according to data characteristics of the communication signal and the
reference signal;
b) responsive to the declared operating state, permitting a main adaptive
filter to acquire data that describe the transfer characteristics of the
medium in at least one signal frequency band;
c) responsive to said data, processing the reference signal to generate an
estimated echo signal; and
d) subtracting said estimated echo signal from the communication signal,
thereby to form an echo-corrected signal; wherein:
e) the declaring step comprises testing for a condition, to be referred to
as "doubletalk," in which the communication signal contains more activity
than can be accounted for solely by the echoes, and further comprises
declaring a particular operating state, to be referred to as the DT state,
whenever doubletalk is detected;
g) permission for the main adaptive filter to acquire new
transfer-characteristic data is withheld during at least the DT state;
h) the step of testing for doubletalk comprises measuring coupling loss
that is associated with the feedback through the medium, measuring energy
of the communication signal, and detecting when said energy exceeds levels
predicted for pure echoes on the basis of said coupling loss; and
i) the step of measuring coupling loss comprises operating an auxiliary
adaptive filter that acquires data describing transfer characteristics of
the medium, detecting when the auxiliary adaptive filter is well adapted,
and updating an estimate of the coupling loss only when the auxiliary
adaptive filter is well adapted.
13. Apparatus for removing, from a communication signal, echoes generated
from a reference signal by feedback through a medium, comprising:
a) means for declaring one of a plurality of operating states, depending on
data characteristics of the communication signal and the reference signal;
b) responsive to the declared operating state, a main adaptive filter for
acquiring data that describe the transfer characteristics of the medium in
at least one signal frequency band;
c) responsive to said data, means for processing the reference signal to
generate an estimated echo signal; and
d) means for subtracting said estimated echo signal from the communication
signal, thereby to form an echo-corrected signal; wherein:
e) the declaring means comprise means for detecting a condition, to be
referred to as "doubletalk, in which the communication signal contains
more activity than can be accounted" for solely by the echoes;
f) the declaring means further comprise means, responsive to the
doubletalk-detection means, for declaring a particular operating state, to
be referred to as the DT state, whenever doubletalk is detected;
g) the apparatus further comprises means for preventing the main adaptive
filter, during at least the DT state, from acquiring new data describing
said transfer characteristics;
h) the apparatus is adapted to permit bidirectional communication during at
least some states which include the DT state; and
i) the doubletalk-detection means comprise: an auxiliary adaptive filter
that is independent of said main adaptive filter and is permitted to
acquire new transfer-characteristic data in all of said operating states,
and means for evaluating the performance of the auxiliary adaptive filter,
said auxiliary adaptive filter acting upon a frequency subband smaller
than, and contained within, the signal frequency band.
14. Apparatus of claim 13, wherein:
the doubletalk-detecting means comprise means for evaluating a statistic,
to be denoted G, that indicates whether the communication signal contains
more energy than the expected echo generated from the reference signal;
and means for evaluating a statistic, to be denoted E, that indicates
whether an echo signal is being effectively removed during the
echo-signal-subtracting step;
the E-evaluating means and the G-evaluating means each operate on
respective signals derived from the communication signal and the reference
signal, within a group of one or more frequency sub-bands that are each
smaller than, and contained within, the signal frequency band;
the declaring means are adapted to declare the DT state only if both G and
E satisfy predetermined conditions;
the E-evaluating means and the G-evaluating means comprise the auxiliary
adaptive filter; and
the auxiliary adaptive filter operates on the sub-band signal or signals
derived from the communication signal.
15. Apparatus of claim 14, wherein, if the reference signal is denoted F,
the communication signal is denoted S, and the echo-corrected signal is
denoted T, then the G-evaluating means further comprise:
a) means for evaluating the statistic L'.sub.a, defined as
##EQU14##
wherein the maxima are taken over appropriate time windows; b) means for
evaluating the statistic L.sub.a, defined as
##EQU15##
wherein the maxima are taken over appropriate time windows; c) a gated
averager for evaluating the statistic L.sub.a, defined as a gated average
of L.sub.a that counts only those values of L.sub.a that occur when E
exceeds a predetermined threshold; and
d) means for setting G=1 if L'.sub.a, >T.sub.G L.sub.a, wherein T.sub.G is
a predetermined threshold, and for setting G=0 otherwise.
16. Apparatus for removing, from a communication signal, echoes generated
from a reference signal by feedback through a medium, comprising:
a) means for declaring one of a plurality of operating states, depending on
data characteristics of the communication signal and the reference signal;
b) responsive to the declared operating state, an adaptive filter for
acquiring data that describe the transfer characteristics of the medium in
at least one signal frequency band;
c) responsive to said data, means for processing the reference signal to
generate an estimated echo signal;
d) means for subtracting said estimated echo signal from the communication
signal, thereby to form an echo-corrected signal; and
e) means for delaying the communication signal and the reference signal
such that each said signal is used as input for said means for declaring
before it is used as input for said adaptive filter and such that the
declaration of the operating state is made before the generation of the
estimated echo signal, wherein:
f) the declaring means comprise means for detecting a condition, to be
referred to as "doubletalk," in which the communication signal contains
more activity than can be accounted for solely by the echoes;
g) the declaring means further comprise means, responsive to the
doubletalk-detection means, for declaring a particular operating state, to
be referred to as the DT state, whenever doubletalk is detected;
h) the apparatus further comprises means for preventing the adaptive
filter, during at least the DT state, from acquiring acquire new data
describing said transfer characteristics; and
i) the apparatus is adapted to permit bidirectional communication during at
least some states which include the DT state.
17. Apparatus of claim 13, or 16, wherein:
the doubletalk-detecting means comprise means for evaluating a statistic,
to be denoted G, that indicates whether the communication signal contains
more energy than the expected echo generated from the reference signal;
and means for evaluating a statistic, to be denoted E, that indicates
whether an echo signal is being effectively removed during the
echo-signal-subtracting step; and
the declaring means are adapted to declare the DT state only if both G and
E satisfy predetermined conditions.
18. Apparatus of claim 17 wherein the E-evaluating means and the
G-evaluating means each operate on respective signals derived from the
communication signal and the reference signal, within a group of one or
more frequency sub-bands that are each smaller than, and contained within,
the signal frequency band.
19. Apparatus of claim 18, wherein said group of one or more frequency
subbands collectively spans a frequency range that is smaller than, and
contained within, the signal frequency band.
20. A communication device adapted for receiving electrical signals to be
referred to as reference signals, said device comprising: a loudspeaker
for transducing the reference signals into acoustic signals; a microphone
for transducing locally generated speech into electrical signals to be
referred to as communication signals; and an echo canceller for removing
from the communication signals echoes generated by acoustic feedback of
the reference signals, thereby to produce an echo-corrected signal,
wherein the echo canceller comprises:
a) means for declaring one of a plurality of operating states, depending on
data characteristics of the communication signals and the reference
signals;
b) responsive to the declared operating state, a main adaptive filter for
acquiring data that describe the transfer characteristics of at least the
acoustic path from the loudspeaker to the microphone in at least one
signal frequency band;
c) responsive to said data, means for processing the reference signals to
generate estimated echo signals; and
d) means for subtracting said estimated echo signals from the communication
signals, thereby to form echo-corrected signals; wherein:
e) the declaring means comprise means for detecting a condition, to be
referred to as "doubletalk," in which a communication signal contains more
activity than can be accounted for solely by the echoes;
f) the declaring means further comprise means, responsive to the
doubletalk-detection means, for declaring a particular operating state, to
be referred to as the DT state, whenever doubletalk is detected;
g) the echo canceller further comprises means for preventing the main
adaptive filter, during at least the DT state, from acquiring new data
describing said transfer characteristics;
h) the echo canceller is adapted to permit bidirectional communication
during at least some states which include the DT state; and
i) the doubletalk-detection means comprise: an auxiliary adaptive filter
that is independent of said main adaptive filter and is permitted to
acquire new transfer-characteristic data in all of said operating states,
and means for evaluating the performance of the auxiliary adaptive filter.
21. Apparatus of claim 20, further comprising delay means for delaying the
communication signals and the reference signals, such that each said
signal is used as input for said means for declaring before it is used as
input for said main adaptive filter and such that the declaration of the
operating state is made before the generation of the estimated echo
signals.
22. Apparatus of claim 20, wherein:
the declaring means further comprise means for detecting a condition
wherein a communication signal substantially contains only an echo of the
reference signal, and means, responsive to detection of such a condition,
for declaring a further operating state, to be referred to as the RX
state;
the communication device further comprises means for determining a value
for a transfer coefficient .sigma..sub.N, and means for applying
.sigma..sub.N to the echo-corrected signal during the RX state; and
the transfer-coefficient-determining means comprise: means for evaluating a
statistic that indicates whether an echo is being effectively removed
during operation of the echo canceller; and means, responsive to at least
certain values of said statistic, for varying .sigma..sub.N continuously
according to the value of said statistic.
23. In a communication system having a microphone and a loudspeaker so
disposed that there is a feedback path from the loudspeaker to the
microphone, apparatus for measuring coupling loss that inheres between the
loudspeaker and the microphone, said loss optionally including gain due to
amplification of input to the loudspeaker and output from the microphone,
the apparatus comprising:
a) an adaptive filter that in operation acquires data that represent
coupling characteristics of the feedback path;
b) means for indicating when the adaptive filter is well adapted; and
c) means, responsive to the indicating means, for updating an estimate of
the coupling loss only when the adaptive filter is well adapted, said
means comprising means for enabling a gated average to be updated when an
indicator of adaptation of said adaptive filter exceeds a threshold, and
for denying said average to be updated at other times.
24. Apparatus for removing, from a communication signal, echoes generated
from a reference signal by feedback through a medium, comprising:
a) means for declaring one of a plurality of operating states, depending on
data characteristics of the communication signal and the reference signal;
b) responsive to the declared operating state, a main adaptive filter for
acquiring data that describe the transfer characteristics of the medium in
at least one signal frequency band;
c) responsive to said data, means for processing the reference signal to
generate an estimated echo signal; and
d) means for subtracting said estimated echo signal from the communication
signal, thereby to form an echo-corrected signal; wherein:
e) the declaring means comprise means for detecting a condition, to be
referred to as "doubletalk," in which the communication signal contains
more activity than can be accounted for solely by the echoes;
f) the declaring means further comprise means, responsive to the
doubletalk-detection means, for declaring a particular operating state, to
be referred to as the DT state, whenever doubletalk is detected;
g) the apparatus further comprises means for preventing the main adaptive
filter, during at least the DT state, from acquiring new data describing
said transfer characteristics;
h) the doubletalk-detection means comprise means for measuring coupling
loss that is associated with the feedback through the medium, means for
measuring energy of the communication signal, and means for detecting when
said energy exceeds levels predicted for pure echoes on the basis of said
coupling loss; and
i) the coupling-loss-measuring means comprise an auxiliary adaptive filter
that in operation acquires data that describe transfer characteristics of
the medium, means for indicating when the adaptive filter is well adapted;
and means, responsive to the indicating means, for updating an estimate of
the coupling loss only when the adaptive filter is well adapted. |
|
 |
|
 |
Claims |
|
