Adaptive echo cancellation

• US 9,602,922 B1
• Filed: 06/27/2013
• Issued: 03/21/2017
• Est. Priority Date: 06/27/2013
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Yang, Jun
• Primary Examiner(s): Goins, Davetta W
• Assistant Examiner(s): SELLERS, DANIEL R
• Application Number:
  US13/929,588
• Time in Patent Office: 1,363 Days
• Field of Search: 381/66
• US Class Current: 1/1
• CPC Class Codes: H04B 3/237 using two adaptive filters, e.g. for near end and for end echo cancelling
• CPC Class Codes: H04M 1/6033 for providing handsfree use or a loudspeaker mode in telephone sets arrangements for preventing acoustic echo H04M9/08
• CPC Class Codes: H04M 3/002 Applications of echo suppressors or cancellers in telephonic connections in two-way loud-speaking telephone systems H04M9/08, echo suppressors or cancellers per se H04B3/20
• CPC Class Codes: H04M 9/082 using echo cancellers echo cancellers per se H04B3/23
• CPC Class Codes: H04R 2499/11 Transducers incorporated or for use in hand-held devices, e.g. mobile phones, PDA's, camera's
• CPC Class Codes: H04R 3/002 Damping circuit arrangements for transducers, e.g. motional feedback circuits
• CPC Class Codes: H04R 3/005 for combining the signals of two or more microphones specially adapted for hearing aids H04R25/407
• CPC Class Codes: H04R 3/02 for preventing acoustic reaction , i.e. acoustic oscillatory feedback specially adapted for hearing aids H04R25/453

Abstract

An audio-based system may perform echo cancellation using first and second adaptive filters. An adaptation controller may be configured to control whether and/or how the filter coefficients of the first adaptive filter are updated, based on detected filter divergence, echo path changes, and/or presence of near-end user voice. The parameters of the second adaptive filter may be copied from the first adaptive filter under certain conditions that indicate whether copying the parameters would be likely to improve echo suppression.

Claims

1. A system, comprising:

a speaker;

a microphone that captures sound from an environment including the speaker and generates an input audio signal;

a first adaptive filter configured to dynamically update a plurality of first filter parameters and to produce a first estimated echo signal representing an echo component of an input audio signal, wherein the echo component is a result of audio propagating over one or more echo paths between the speaker and the microphone;

an adaptation controller configured to;

reset the first filter parameters to default;

determine that the first adaptive filter is converging;

determine that the one or more echo paths are stable;

determine that a near-end user voice is detected in the input audio signal;

stop the dynamically updating of the first filter;

a second adaptive filter configured to

receive the first filter parameters from the first adaptive filter, and

produce a second estimated echo signal representing the echo component of the input audio signal based at least in part on the first filter parameters;

an echo path change detection component configured to detect a change in the one or more echo paths based, at least in part, on one or more of a change in position of a user relative to the microphone, or a change in position of the speaker;

a parameter update controller configured to update second filter parameters from the first filter parameters under one or more specified conditions, wherein at least one of the one or more specified conditions is based, at least in part, on the change in the one or more echo paths; and

a subtraction component configured to subtract the second estimated echo signal from the input audio signal to produce an echo-suppressed audio signal.

2. The computing device of claim 1, wherein the first filter parameters are updated at an update rate, and wherein the adaptation controller is further configured to increase the update rate when both of the following conditions exist:

(a) the first adaptive filter is not diverging and (b) one or more of the echo paths are changing.

3. The computing device of claim 1, further comprising evaluating the one or more specified conditions are based at least in part on one or more of the following:

energy of the input audio signal;

energy of an audio reference signal;

energy of the echo-suppressed audio signal;

energy of the first estimated echo signal;

or

energy of the second estimated echo signal.

4. The computing device of claim 1, wherein the one or more specified conditions are determined by:

comparing an energy of the input audio signal to a first threshold;

comparing an energy of an audio reference signal to a second threshold;

comparing a relationship between an energy of the echo suppressed audio signal and an energy of an error signal to a third threshold; and

comparing a relationship between the energy of the input audio signal and the energy of the error signal to a fourth threshold;

wherein the error signal comprises a difference between the input audio signal and the first estimated echo signal.

5. A method, comprising:

receiving an input audio signal having an echo component, wherein the echo component is a result of audio propagating over one or more echo paths between a speaker and a microphone;

suppressing the echo component of the input audio signal to produce an echo-suppressed audio signal, wherein the suppressing is based at least in part on a plurality first filter parameters that are dynamically updated;

detecting whether the first filter parameters are diverging;

detecting whether a change has occurred in the one or more echo paths;

detecting whether near-end audio is represented by the input audio signal;

detecting that the first filter parameters are diverging;

resetting the first filter; and

determining that the one or more echo paths are stable based, at least in part, on one or more of a position of a user relative to the microphone, a change in position of the microphone, or a change in position of the speaker;

stopping the dynamic updating of the first filter parameters.

6. The method of claim 5, further comprising estimating a first echo signal based at least in part on the first filter parameters and stopping the dynamic updating of the first filter parameters in response to detecting near-end audio represented by the input audio signal, the detecting the near-end audio represented by the input audio signal based at least in part on power of the input audio signal, power of the estimated first echo signal, and power of an error signal, the error signal comprises a difference between the input audio signal and the estimated first echo signal.

7. The method of claim 5, further comprising estimating a first echo signal based at least in part on the first filter parameters, wherein detecting whether a change has occurred in the one or more echo paths is based at least in part on power of the input audio signal and power of an error signal, and wherein the error signal comprises a difference between the input audio signal and the estimated first echo signal.

8. The method of claim 5, further comprising estimating a first echo signal with an adaptive finite impulse response filter by minimizing a power of an error signal, wherein the error signal comprises a difference between the input audio signal and the estimated first echo signal.

9. The method of claim 5, wherein the first filter parameters are updated at an update rate, the method further comprising increasing the update rate if the first filter parameters are diverging.

10. The method of claim 5, wherein the suppressing comprises:

estimating an echo signal corresponding to the echo component of the input audio signal based at least in part on second filter parameters; and

subtracting the estimated second echo signal from the input audio signal to produce the echo-suppressed audio signal.

11. The method of claim 10, wherein estimating the echo signal comprises filtering a received reference audio signal with an adaptive finite impulse respond filter to produce the echo signal.

12. The method of claim 5, wherein the suppressing comprises:

estimating a first echo signal corresponding to the echo component of the input audio signal based at least in part on the first filter parameters;

estimating a second echo signal corresponding to the echo component of the input audio signal based at least in part on second filter parameters;

updating the second filter parameters by copying the first filter parameters under one or more specified conditions; and

subtracting the second estimated echo signal from the input audio signal to produce the echo-suppressed audio signal.

13. The method of claim 12, further comprising evaluating the one or more specified conditions based at least in part on one or more of the following:

energy of the input audio signal;

energy of an output audio signal;

energy of the echo-suppressed audio signal;

energy of the estimated first echo signal;

or

energy of the estimated second echo signal.

14. The method of claim 12, further comprising evaluating the one or more specified conditions by:

comparing an energy of the input audio signal to first threshold;

comparing an energy of an output reference signal to a second threshold, wherein estimating the first and second echo signals is based at least in part on the output reference signal;

comparing a relationship between an energy of the echo-suppressed audio signal and an energy of an error signal to a third threshold;

comparing a relationship between an energy of the input audio signal and the energy of the error signal to a fourth threshold; and

wherein the echo-suppressed audio signal comprises the input audio signal minus the estimated second echo signal and the error signal comprises the input audio signal minus the estimated first echo signal.

15. A system, comprising:

one or more processors;

a microphone;

a speaker;

memory storing computer-executable instructions that, when executed by one or more processors, cause the one or more processors to perform acts comprising;

estimating a first echo signal representing an echo component of an input audio signal by dynamically updating a plurality of first filter parameters and applying the first filter parameters to an audio reference signal, wherein the echo component is a result of audio propagating over an echo path between the speaker and the microphone;

estimating a second echo signal representing the echo component of the input audio signal by applying a plurality of second filter parameters to the audio reference signal; and

updating the second filter parameters from the first of filter parameters under one or more specified conditions, wherein at least one of the one or more specified conditions is based, at least in part, on a change in position of a user relative to the microphone, or a change in position of the speaker.

16. The system of claim 15, wherein estimating the first echo signal and estimating the second echo signal are performed by first and second finite impulse response filters, respectively.

17. The system of claim 15, the acts further comprising evaluating the one or more specified conditions based at least in part on one or more of the following:

energy of the input audio signal;

energy of the audio reference signal;

energy of the estimated first echo signal;

or

energy of the estimated second echo signal.

18. The system of claim 15, the acts further comprising evaluating the one or more specified conditions by:

comparing an energy of the input audio signal to first threshold;

comparing an energy of the audio reference signal to a second threshold;

comparing a relationship between an energy of an echo suppressed audio signal and an energy of an error signal to a third threshold; and

comparing a relationship between an energy of the input audio signal and the energy of the error signal to a fourth threshold;

wherein the echo-suppressed audio signal comprises the input audio signal minus the estimated second echo signal and the error signal comprises the input audio signal minus the estimated first echo signal.

19. The system of claim 15, the acts further comprising:

stopping the dynamic updating of the first filter parameters in response to divergence of the first filter parameters; and

stopping the dynamic updating of the first filter parameters when both of the following conditions exist;

(a) echo path changes have not occurred, and (b) near-end audio is represented by the input audio signal.

20. The system of claim 15, the acts further comprising:

resetting the first filter parameters in response to divergence of the first filter parameters; and

stopping the dynamic updating of the first filter parameters when all of the following conditions exist;

(a) the first filter parameters are not diverging, (b) echo path changes have not occurred, and (c) near-end audio is represented by the input audio signal.