Echo Cancellers and Echo Cancelling Methods

• US 20110311066A1
• Filed: 06/16/2010
• Published: 12/22/2011
• Est. Priority Date: 06/16/2010
• Status: Active Grant

• Current Assignee: Lectrosonics, Inc.
• Original Assignee: Lectrosonics, Inc.
• Inventors: Nagy, Ferenc
• Granted Patent: US 8,619,970 B2
• Time in Patent Office: Days
• US Class Current: 381/66
• CPC Class Codes: H04B 3/23 using a replica of transmitted signal in the time domain, e.g. echo cancellers

Abstract

Methods and systems for updating the adaptive filter of an echo canceller. A method of updating the adaptive filter of an echo canceller in which an estimated echo is resolved from a received signal and then subtracted from an incoming echo-contaminated signal so as to produce a filtered output signal, includes: obtaining a corrected impulse response of an echo reconstruction filter (ERF); calculating specified decision measures usable to decide whether to prospectively apply the corrected ERF impulse response or a current ERF impulse response; determining whether application of the corrected ERF impulse response would result in improved echo cancellation; and updating the ERF to apply the corrected impulse response, when it is determined that the updating would result in a lower residual echo.

Claims

1. A method of updating the adaptive filter of an echo canceller in which an estimated echo is resolved from a received signal and then subtracted from an incoming echo-contaminated signal so as to produce a filtered output signal, comprising:

obtaining a corrected impulse response (hc) of an echo reconstruction filter (ERF);

calculating specified decision measures usable to decide whether to prospectively apply the corrected ERF impulse response hc or a current ERF impulse response he to a reference signal (X), the decision measures being

(i) a mutual information rate (infoVeX) of an estimate of the signal plus noise (Ve) and the reference signal (X),

(ii) an estimated echo return loss enhancement (ERLEi) of the echo canceller, and

(iii) a corrected echo return loss enhancement (ERLEc) of the echo canceller;

determining whether application of the corrected ERF impulse response hc would result in improved echo cancellation by

(i) comparing a mutual information rate of infoVeX to a specified threshold, and

(ii) comparing the ERLEc to the ERLEi; and

updating the ERF to apply the corrected impulse response hc, when it is determined that the updating would result in a lower residual echo (Er), based on a comparison of past performance of the ERF using the corrected impulse response hc and past performance of the ERF using the current ERF impulse response he.

2. The method of claim 1, wherein, when infoVeX is not greater than the threshold and ERLEc is greater than ERLEi, the ERF is updated with the corrected ERF impulse response hc.

3. The method of claim 1, wherein, when infoVeX is greater than the threshold, the ERLEe is set to the ERLEi, and if ERLEc is greater than ERLEi the ERF is updated with the corrected ERF impulse response hc.

4. The method of claim 1, further comprising saving the ERLEc for future updates, when the ERF is updated.

5. The method of claim 1, wherein the ERLEi is based on a currently active ERF impulse response, and the ERLEc is based on results that would have occurred if the corrected ERF impulse response had been applied.

6. The method of claim 1, wherein, in the determining, a record of past observations of the echo canceller is used and an actual output of the echo canceller is not used.

7. The method of claim 1, wherein the obtaining comprises:

calculating a record of the estimated echo (Ee) from the ERF using a current ERF impulse response he and a record of a reference signal X;

resolving an estimate of the local signal plus noise Ve_n by subtracting the estimated echo Ee from a record of an input signal Y_n.

estimating a residual frequency response Hr with a record of Ve and X using spectral estimates of S_X and S_Ve,X; and

computing the corrected ERF impulse response (hc) based on the spectral estimates, as follows;

8. The method of claim 7, wherein, in the calculating a record of the estimated echo Ee, a fast convolution algorithm is used and single segment or multiple segments are used.

9. The method of claim 8, wherein, the convergence rate and the depth of the final echo cancellation depend on selected ones of a segment length, an overlapping ratio, and a window (tapering) function.

10. The method of claim 9, wherein the segment length is 8/3 L, L is the tail length of the ERF in samples, the overlapping ratio is a 50% overlapping ratio, and the window function is a Hann window function.

13. The echo canceller of claim 8, wherein the logic section is a fixed-point DSP, a floating point DSP, or an FPGA.

11. A method of updating an echo reconstruction filter (ERF) having changeable coefficients, comprising:

calculating a corrected impulse response (hc) of an echo reconstruction filter (ERF);

calculating specified decision measures usable to decide whether to apply a current ERF or an updated ERF to echo-contaminated signals in the future;

deciding, based on the calculated decision measures, whether the hc would have yielded a lower residual echo (Er) than the current impulse response of the ERF (he) when applied to past observations; and

replacing the coefficients of the ERF so that an impulse response of the ERF (he) equals hc.

12. An echo canceller that cancels an echo from an echo-contaminated signal, comprising:

a logic section that

applies an echo reconstruction filter (ERF) to a reference signal (X) to resolve an estimated echo (Ee), the ERF based on an adaptive filter with coefficients that are changeable based on a peak echo-to-noise ratio of the echo-contaminated signal,

determines a corrected impulse response (hc) of an echo reconstruction filter (ERF),

evaluates whether a corrected echo reconstruction filter (ERFc) with coefficients that cause an impulse response thereof to match hc would have resulted in a lower residual echo (Er) than those of the current ERF when applied to past observations, and

replaces the coefficients so that the impulse response of the ERF (he) matches hc, when the hc would have resulted in a lower residual echo; and

a subtractor that filters the echo from the echo-contaminated signal (Y) by subtracting the resolved Ee therefrom to produce a filtered output signal (Ve).

14. An echo cancelling method, comprising:

applying, to an echo-contaminated signal, an adaptive filter with coefficients that are changeable based on a peak echo-to-noise ratio of the echo-contaminated signal;

maintaining at least a partial record of echo cancellation performance of the applying;

determining a corrected echo reconstruction filter (ERFc);

deciding whether the ERFc will result in improved echo cancellation, based on comparisons of echo return loss enhancement (ERLE) of past echo canceller performance using a current ERF and the ERFc;

replacing the ERF with the ERFc when the ERFc would have resulted in a lower residual echo Er than the ERF; and

subtracting an estimated echo (Ee) resolved by the ERF from future incoming echo-contaminated signals.