Post-mixing acoustic echo cancellation systems and methods
Acoustic echo cancellation systems and methods are provided that can cancel and suppress acoustic echo from the output of a mixer that has mixed audio signals from a plurality of acoustic sources, such as microphones. The microphones may have captured speech and sound from a remote location or far end, such as in a conferencing environment. The acoustic echo cancellation may generate an echo-cancelled mixed audio signal based on a mixed audio signal from a mixer, information gathered from the audio signal from each of the plurality of acoustic sources, and a remote audio signal. The systems and methods may be computationally efficient and resource-friendly.
1. A system, comprising:
(A) a memory;
(B) a plurality of acoustic sources each configured to generate an audio signal;
(C) a mixer in communication with the plurality of acoustic sources and the memory, the mixer configured to mix the audio signal from each of the plurality of acoustic sources to produce a mixed audio signal; and
(D) an acoustic echo canceller in communication with the mixer, the memory, and a remote audio signal, the acoustic echo canceller configured to generate an echo-cancelled mixed audio signal based on the mixed audio signal, information gathered from the audio signal from each of the plurality of acoustic sources, and the remote audio signal, wherein the acoustic echo canceller comprises:
a background filter having background filter tap coefficients and configured to measure a background error power of the audio signal from each of the plurality of acoustic sources using a normalized least-mean squares algorithm;
a hidden filter having hidden filter tap coefficients and configured to measure a hidden error power of the audio signal from each of the plurality of acoustic sources, based on the audio signal from each of the plurality of acoustic sources and the remote audio signal; and
an error comparison module in communication with the background filter and the hidden filter, the error comparison module configured to:
compare the background error power and the hidden error power; and
select and store the background filter tap coefficients in the memory, if the background error power is greater than the hidden error power.
2. The system of claim 1 :
further comprising a signal selection mechanism in communication with the plurality of acoustic sources and the acoustic echo canceller, the signal selection mechanism configured to select at least one audio signal from at least one of the plurality of acoustic sources and convey the at least one selected audio signal to the acoustic echo canceller;
wherein the acoustic echo canceller is further configured to generate the echo-cancelled mixed audio signal based on the mixed audio signal, information gathered from the at least one selected audio signal, and the remote audio signal.
3. The system of claim 1 , wherein the error comparison module is further configured to copy the stored background filter tap coefficients from the memory to replace the hidden filter tap coefficients, if the background error power is greater than the hidden error power.
4. The system of claim 1 , wherein the background filter is configured to measure a background error e[n] according to the equation:
e [ n ]= d [ n ]− ĥ † [ n ] x [ n ]
where d[n] is one of the audio signals, x[n] is a vector of samples from the remote audio signal, and † denotes a conjugate transpose operation;
wherein the background error power is estimated based on the background error.
5. The system of claim 1 , wherein the error comparison module is further configured to update the background filter tap coefficients according to the equation:
h
^
[
n
+
1
]
=
h
^
[
n
]
+
α
x
[
n
]
2
e
*
[
n
]
x
[
n
]
where α is a step-size parameter, * denotes a complex conjugation operation, and ∥⋅∥ denotes a l 2 norm.
6. The system of claim 1 , wherein the acoustic echo canceller further comprises:
a mix filter having mix filter tap coefficients and tap weights, and configured to filter the remote audio signal to generate a filtered remote audio signal.
7. The system of claim 1 , wherein:
the acoustic echo canceller further comprises a mix filter having mix filter tap coefficients and tap weights, and configured to filter the remote audio signal to generate a filtered remote audio signal; and
the error comparison module is further configured to copy the stored background filter tap coefficients from the memory to update the mix filter tap coefficients by combining the hidden filter tap coefficients of each of the plurality of acoustic sources not currently under adaptation and the most recently updated background filter tap coefficients corresponding to the acoustic source currently under adaptation, if the background error power is greater than the hidden error power.
8. The system of claim 1 , wherein:
the acoustic echo canceller further comprises a mix filter having mix filter tap coefficients and tap weights, and configured to filter the remote audio signal to generate a filtered remote audio signal; and
the mix filter is further configured to be updated if a channel scaling factor of the mixer has changed by updating the tap weights corresponding to the changed channel scaling factor by adding the difference in weight multiplied by a channel impulse response estimate.
9. The system of claim 6 , wherein the acoustic echo canceller is configured to generate the echo-cancelled mixed audio signal by subtracting the filtered remote audio signal from the mixed audio signal.
10. The system of claim 9 , wherein the acoustic echo canceller further comprises:
a mix estimator in communication with the mixer, the mix filter, and the echo-cancelled mixed audio signal, the mix estimator configured to:
measure an output coherence of the filtered remote audio signal; and
estimate a residual echo power of the echo-cancelled mixed audio signal; and
a non-linear processor configured to process the echo-cancelled mixed audio signal to generate an echo-suppressed mixed audio signal, if the output coherence exceeds a predetermined threshold or if the residual echo power exceeds half of an power of the mixed audio signal.
11. A method, comprising:
receiving an audio signal from each of a plurality of acoustic sources;
receiving a remote audio signal;
mixing the audio signal from each of the plurality of acoustic sources using a mixer to produce a mixed audio signal; and
generating an echo-cancelled mixed audio signal based on the mixed audio signal, information gathered from the audio signal from each of the plurality of acoustic sources, and the remote audio signal, using an acoustic echo canceller, wherein generating the echo-cancelled mixed audio signal comprises:
measuring a background error power of the audio signal from each of the plurality of acoustic sources using a normalized least-mean squares algorithm in a background filter having background filter tap coefficients;
measuring a hidden error power of the audio signal from each of the plurality of acoustic sources, based on the audio signal from each of the plurality of acoustic sources and the remote audio signal, using a hidden filter having hidden filter tap coefficients;
comparing the background error power and the hidden error power; and
selecting and storing the background filter tap coefficients in a memory, if the background error power is greater than the hidden error power.
12. The method of claim 11 :
further comprising selecting and conveying at least one selected audio signal from at least one of the plurality of acoustic sources using a signal selection mechanism to the acoustic echo canceller;
wherein generating the echo-cancelled mixed audio signal comprises generating the echo-cancelled mixed audio signal based on the mixed audio signal, information gathered from the at least one selected audio signal, and the remote audio signal.
13. The method of claim 11 , further comprising copying the stored background filter tap coefficients from the memory to replace the hidden filter tap coefficients, if the background error power is greater than the hidden error power.
14. The method of claim 11 , wherein:
measuring the background error power comprises measuring a background error e[n] according to the equation:
e [ n ]= d [ n ]− ĥ † [ n ] x [ n ]
where d[n] is one of the audio signals, x[n] is a vector of samples from the remote audio signal, and † denotes a conjugate transpose operation; and
estimating the background error power based on the background error.
15. The method of claim 11 , further comprising updating the background filter tap coefficients according to the equation:
h
^
[
n
+
1
]
=
h
^
[
n
]
+
α
x
[
n
]
2
e
*
[
n
]
x
[
n
]
where α is a step-size parameter, * denotes a complex conjugation operation, and ∥⋅∥ denotes a l 2 norm.
16. The method of claim 11 , further comprising filtering the remote audio signal to generate a filtered remote audio signal using a mix filter having mix filter tap coefficients and tap weights.
17. The method of claim 11 , further comprising:
filtering the remote audio signal to generate a filtered remote audio signal using a mix filter having mix filter tap coefficients and tap weights; and
copying the stored background filter tap coefficients from the memory to update the mix filter tap coefficients by combining the hidden filter tap coefficients of each of the plurality of acoustic sources not currently under adaptation and the most recently updated background filter tap coefficients corresponding to the acoustic source currently under adaptation, if the background error power is greater than the hidden error power.
18. The method of claim 11 , further comprising:
filtering the remote audio signal to generate a filtered remote audio signal using a mix filter having mix filter tap coefficients and tap weights; and
if a channel scaling factor of the mixer has changed, updating the mix filter by updating the tap weights corresponding to the changed channel scaling factor by adding the difference in weight multiplied by a channel impulse response estimate.
19. The method of claim 16 , wherein generating the echo-cancelled mixed audio signal comprises subtracting the filtered remote audio signal from the mixed audio signal.
20. The method of claim 19 , further comprising:
measuring an output coherence of the filtered remote audio signal using a mix estimator;
estimating a residual echo power of the echo-cancelled mixed audio signal, using the mix estimator; and
processing the echo-cancelled mixed audio signal to generate an echo-suppressed mixed audio signal using a non-linear processor, if the output coherence exceeds a predetermined threshold or if the residual echo power exceeds half of an power of the mixed audio signal.