CROSS-CORRELATION BASED ECHO CANCELLER CONTROLLERS

• US 20080240413A1
• Filed: 04/02/2007
• Published: 10/02/2008
• Est. Priority Date: 04/02/2007
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Stokes, Jack W., Mohammad, Asif Iqbal, Grant, Steven L.
• Granted Patent: US 8,014,519 B2
• Time in Patent Office: Days
• US Class Current: 379/406.08
• CPC Class Codes: H04M 9/082 using echo cancellers echo cancellers per se H04B3/23

Abstract

Cross-correlation based echo canceller controllers are described herein. By way of example, a system for controlling an echo canceller having one or more adaptive filters can include one or more adaptive filter controllers each corresponding to one of the one or more adaptive filters and each configured to halt adaptation of its corresponding adaptive filter according to the cross-correlation of its corresponding corrupted signal and its corresponding error signal of its corresponding adaptive filter.

Claims

1. A system for controlling an echo canceller having one or more adaptive filters, each of the one or more adaptive filters configured to receive a corrupted signal and to produce an error signal, comprising:

one or more adaptive filter controllers each corresponding to one of the one or more adaptive filters and each configured to halt adaptation of its corresponding adaptive filter according to a cross-correlation of its corresponding corrupted signal and its corresponding error signal of its corresponding adaptive filter.

2. The system of claim 1, wherein the one or more adaptive filter controllers are configured to employ a decision statistic to halt adaptation of their corresponding adaptive filters.

3. The system of claim 2, wherein the decision statistic is where {circumflex over (r)}_em is

$1-\frac{{\hat{r}}_{\mathrm{em}}}{{\stackrel{}{\sigma }}_{}^{}}$

^ m 2 ,

based on an estimate of a time domain cross-correlation coefficient between the corrupted signal and the error signal and {circumflex over (σ

)}_m² is based on an estimate of a time domain variance of the corrupted signal.

4. The system of claim 2, wherein the decision statistic is a frequency domain decision statistic.

5. The system of claim 4, wherein the frequency domain decision statistic is

$1-\frac{{\hat{r}}_{\mathrm{em}}}{}$

( k , t ) σ

^ m 2 

( k , t ) ,

where {circumflex over (r)}_em(k,t) is an estimate of a cross-correlation coefficient between the corrrupted signal and the error signal for a kth frequency subband of the tth frame and {circumflex over (σ

)}_m² (k,t) is an estimate of a variance of the corrupted signal for the kth frequency subband of the tth frame.

6. The system of claim 2, wherein the decision statistic comprises a noise term.

7. The system of claim 6, wherein the decision statistic is

$1-\frac{{\hat{r}}_{\mathrm{em}}}{}$

( k , t ) - σ

^ z 2 

( k , t ) σ

^ m 2 

( k , t ) ,

where {circumflex over (r)}_em(k,t) is an estimate of a cross-correlation coefficient between the corrupted signal and the error signal for a kth frequency subband and tth frame, {circumflex over (σ

)}_m²(k,t) is an estimate of a variance of the corrupted signal for the kth frequency subband and tth frame and {circumflex over (σ

)}_z²(k,t) is a variance of the noise in the kth subband and the tth frame.

8. The system of claim 6, wherein the noise term is an estimate of noise in a kth frequency subband of the corrupted signal.

9. The system of claim 2, wherein the decision statistic is an overall decision statistic.

10. The system of claim 9, wherein the overall decision statistic is based on individual decision statistics corresponding to each of the one or more cross-correlation controllers.

11. Computer-executable instructions for performing a method of controlling one or more adaptive filters of an echo canceller, the computer-executable instructions stored on one or more computer-readable media, the method comprising:

receiving a corrupted signal;

receiving one or more cancellation errors from the one or more adaptive filters of the echo canceller;

computing one or more decision statistics based on a cross-correlation of the corrupted signal and the one or more cancellation errors;

comparing the one or more decision statistic to their threshold values; and

controlling the one or more adaptive filters based on the comparison of the one or more decision statistics to their threshold values.

12. The computer-executable instructions of claim 11, wherein computing the one or more decision statistics based on the cross-correlation of the corrupted signal and the one or more cancellation errors comprises computing a plurality of decision statistics each corresponding to one of a plurality of frequency subbands and wherein comparing the decision statistic to the threshold value comprises comparing each of the plurality of decision statistics to its threshold value.

13. The computer-executable instructions of claim 11, wherein the one or more decision statistics are given by

$1-\frac{{\hat{r}}_{\mathrm{em}}}{}$

( k , t ) - σ

^ z 2 

( k , t ) σ

^ m 2 

( k , t ) ,

where {circumflex over (r)}_em(k,t) is an estimate of a cross-correlation coefficient between the corrupted signal and the error signal for a kth frequency subband and tth frame, {circumflex over (σ

)}_m²(k,t) is an estimate of a variance of the corrupted signal for the kth frequency subband and tth frame and {circumflex over (σ

)}_z²(k,t) is a variance of the noise in the kth subband and the tth frame.

14. The computer-executable instructions of claim 11, wherein the one or more decision statistics are given by

$1-\frac{{\hat{r}}_{\mathrm{em}}}{}$

( k , t ) σ

^ m 2 

( k , t ) ,

where {circumflex over (r)}_em(k,t) is an estimate of a cross-correlation coefficient between the corrrupted signal and the error signal for a kth frequency subband of the tth frame and {circumflex over (σ

)}_m²(k,t) is an estimate of a variance of the corrupted signal for the kth frequency subband of the tth frame.

15. The computer-executable instructions of claim 11, wherein controlling the one or more adaptive filters based on the comparison of the one or more decision statistics to their threshold values comprises employing an overall decision statistic.

16. The computer-executable instructions of claim 11, wherein computing the one or more decision statistics based on the cross-correlation of the corrupted signal and the one or more cancellation errors comprises employing an exponential recursive weighting algorithm.

17. A system for facilitating communication, comprising:

means for canceling an echo in a corrupted signal, the means for canceling the echo comprising one or more adaptive filters; and

means for stopping the one or more adaptive filters from adapting based on a cross-correlation of the corrupted signal with a residual signal produced by the means for canceling the echo.

18. The system of claim 17, wherein the means for stopping is implemented in the frequency domain.

19. The system of claim 17, wherein the means for stopping comprises a decision statistic.

20. The system of claim 19, wherein the decision statistic is an overall decision statistic.