Echo cancellation

Patent metadata

Field	Value
Publication number	US-2016309042-A1
Application number	US-201415036963-A
Country	US
Kind code	A1
Filing date	Nov 5, 2014
Priority date	Dec 12, 2013
Publication date	Oct 20, 2016
Grant date	—

Abstract

An audio echo suppressor comprises a first receiver ( 201 ) which receives a first audio signal for rendering by a loud-speaker ( 205 ) and a second receiver ( 211 ) which receives a microphone signal. A linear echo-cancellation filter ( 213 ) generates a first compensation signal from the first audio signal and a compensator ( 215 ) generates a residual signal by compensating the microphone signal for the first compensation signal. A first adapter ( 217 ) determines filter parameters for the linear echo-cancellation filter ( 213 ). An estimator ( 219 ) generates distortion measures where each distortion measure is indicative of the contribution to the residual signal in a frequency interval outside a first frequency interval. The residual signal results from rendering of signal components of the first audio signal that are within the first frequency interval. An echo reducer ( 207 ) then performs echo suppression based on the distortion measures. The approach may provide improved echo suppression, especially of non-linear echoes caused by non-linear effects.

First claim

1 . An audio echo suppressor comprising:
a first receiver for receiving a first audio signal for rendering by a loudspeaker, a second receiver for receiving a microphone signal; a linear echo-cancellation filter for generating a first compensation signal from the first audio signal; a compensator for generating a residual signal by subtracting the first compensation signal from the microphone signal; a first adapter for determining a set of filter parameters for the linear echo-cancellation filter in response to the residual signal; a first estimator arranged to generate a set of at least one distortion measure providing indication of non-linear distortion which is caused by rendering of signal components in a first frequency interval of the first audio signal, said distortion measure being indicative of a contribution to the residual signal in at least one frequency interval that is outside the first frequency interval, and an echo reducer arranged to perform echo suppression in response to the set of at least one distortion measure.

2 . The audio echo suppressor of claim 1 wherein the first estimator comprises:
a first subband generator for generating a first subband signal for the first frequency interval from the first audio signal;
a second subband generator for generating a plurality of residual subband signals from the residual signal;
a comparator for determining a similarity measure for at least some residual subband signals not belonging to the first frequency interval, the similarity measure for a residual subband signal being indicative of a similarity of the residual subband signal and the first subband signal; and
a second estimator for determining the distortion measure for the first frequency interval from the similarity values.

3 . The audio echo suppressor of claim 2 wherein the comparator is arranged to generate the similarity measure for a first residual subband signal in response to a similarity of signal variations of the first subband signal and signal variations of the first residual subband signal.

4 . The audio echo suppressor of claim 2 wherein the comparator is arranged to generate the similarity measure for a first residual subband signal in response to a cross correlation of the first subband signal and the first residual subband signal.

5 . The audio echo suppressor of claim 2 wherein the first estimator is arranged to generate distortion measures for a plurality of frequency intervals, and the echo reducer is arranged to separately adjust parameters of the echo suppression for each frequency interval of the plurality of frequency intervals in response to the distortion measure for the frequency interval.

6 . The audio echo suppressor of claim 2 wherein the second estimator is arranged to generate the distortion measure by combining similarity values for a plurality of the at least some residual subband signals.

7 . The audio echo suppressor of claim 2 wherein the second estimator is arranged to normalize the distortion measure with respect to a power estimate for the contributions to the microphone signal from signal components of the first audio signal within the first frequency interval.

8 . The audio echo suppressor of claim 7 wherein the second estimator is arranged to generate the power estimate in response to a linear echo power estimate for the first frequency interval, the linear echo power estimate being generated in response to the set of filter parameters.

9 . The audio suppressor of claim 2 wherein the comparator is arranged to select the at least some residual subband signals in response to a harmonic relationship between a frequency of the first frequency interval and a subband frequency of the residual subband signals.

10 . The audio echo suppressor of claim 1 wherein the first estimator comprises:
a first subband generator for generating a first residual subband signal for the first frequency interval from the residual signal;
a second subband generator for generating a plurality of first audio subband signals from the first audio signal;
a comparator for determining a similarity measure for at least some first audio subband signals not belonging to the first frequency interval, the similarity measure for a first audio subband signal being indicative of a similarity of the first audio subband signal and the first residual subband signal; and
a second estimator for determining the distortion measure for the first frequency interval from the similarity values.

11 . The audio echo suppressor of claim 10 wherein the second estimator is arranged to generate the distortion measure for the first residual subband signal based on a combination of similarity values for a plurality of the at least some first audio subband signals.

12 . The audio echo suppressor of claim 1 wherein the echo reducer is arranged to modify the first audio signal in response to the set of distortion measures prior to rendering by the loudspeaker.

13 . The audio echo suppressor of claim 12 wherein the echo reducer is arranged to set a gain or a clipping level for the first audio signal in the first frequency interval in response to the distortion measure for the first frequency interval.

14 . The audio echo suppressor of claim 1 wherein the echo reducer is arranged to modify the residual signal in response to the set of distortion measures.

15 . A method of audio echo suppression comprising:
receiving a first audio signal for rendering by a loudspeaker, receiving a microphone signal; a linear echo-cancellation filter generating a first compensation signal from the first audio signal; generating a residual signal by subtracting the first compensation signal from the microphone signal; determining a set of filter parameters for the linear echo-cancellation filter in response to the residual signal; generating a set of at least one distortion measure providing indication of non-linear distortion which is caused by rendering of signal components in a first frequency interval of the first audio signal, said distortion measure being indicative of a contribution to the residual signal in at least one frequency interval that is outside the first frequency interval; and adapting an echo suppression in response to the set of at least one distortion measure.

Assignees

• Koninklijke Philips Nv

Inventors

• Kechichian Patrick
• Janse Cornelis Pieter

Classifications

• H04M9/085 Primary

  using digital techniques (H04M9/082 takes precedence) · CPC title

• G10K11/175

  using interference effects; Masking sound · CPC title