Audio echo cancellation with robust double-talk detection in a conferencing environment

• US 10,154,148 B1
• Filed: 08/03/2017
• Issued: 12/11/2018
• Est. Priority Date: 08/03/2017
• Status: Active Grant

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Polycom Incorporated (Plantronics, Inc.)
• Inventors: Chu, Peter L., Elias, Eric
• Primary Examiner(s): Gauthier, Gerald
• Application Number:
  US15/667,910
• Time in Patent Office: 495 Days
• Field of Search: 348 1408, 370260, 370289, 37940601, 37940608, 37940612, 37940609, 381 59, 381 66, 381 711, 381104, 381315, 381321, 381356, 381 93, 708313
• CPC Class Codes: G10L 2021/02082 the noise being echo, reverberation of the speech
• CPC Class Codes: G10L 21/0208 Noise filtering
• 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 3/567 Multimedia conference systems
• CPC Class Codes: H04M 3/568 audio processing specific to telephonic conferencing, e.g. spatial distribution, mixing of participants echo suppression in two-way loud-speaking telephone systems H04M9/02; sound field processing per se H04S7/30
• CPC Class Codes: H04M 9/082 using echo cancellers echo cancellers per se H04B3/23
• CPC Class Codes: H04N 7/147 Communication arrangements, e.g. identifying the communication as a video-communication, intermediate storage of the signals selecting H04Q
• CPC Class Codes: H04N 7/15 Conference systems

Abstract

A conferencing endpoint includes a loudspeaker, a base microphone, and a double-talk detection module which allows two-way communication between the conferencing endpoint and a remote endpoint only when participants at both endpoints are speaking at the same time, so as to minimize echo due to feedback. The double-talk detection module adds the energy of any distortion from the loudspeaker to the energy of the signal coming from the remote endpoint, and compares this combined energy with the energy of the base microphone to determine whether double-talk is present. The double-talk detection module is thus prevented from mistaking the feedback for near end talk at the endpoint.

Claims

1. A method of preventing false positives by a double-talk detection unit at a conferencing endpoint, the method comprising:

receiving a first signal;

determining an energy value of the first signal;

emitting audio at a loudspeaker, the audio based on the first signal;

collecting audio at a first microphone, the audio including a first linear component corresponding to the first signal, and a first non-linear component corresponding to distortion of the first signal within the emitted audio;

emitting, by the first microphone, a first microphone signal, the first microphone signal comprising a first linear portion corresponding to the first linear component of the collected audio and a non-linear portion corresponding to the first non-linear component of the collected audio;

determining an energy value associated with the non-linear portion of the first microphone signal;

transmitting an energy signal to a double-talk detection unit of a second microphone, the energy signal corresponding to the energy value of the non-linear portion of the first microphone signal multiplied by a scaling factor;

capturing audio at the second microphone, the audio including a second linear component corresponding to the first signal, and a second non-linear component corresponding to distortion of the first signal within the emitted audio, wherein the second linear component is attenuated relative the first linear component, and the second non-linear component is attenuated relative the first non-linear component;

determining an energy value of the audio captured at the second microphone;

receiving the transmitted energy signal at the double-talk detection unit;

calculating, by the double-talk detection unit, a sum of the energy value of the non-linear portion of the first microphone signal multiplied by the scaling factor with the energy value of the first signal; and

comparing, by the double-talk detection unit, the sum with the energy value of the audio captured at the second microphone, whereby the double-talk detection unit is prevented from falsely detecting double-talk.

2. The method of claim 1, wherein determining the energy value associated with the non-linear portion of the first microphone signal comprises isolating the non-linear portion of the first microphone signal.

3. The method of claim 2, wherein isolating the non-linear portion of the first microphone signal comprises subtracting the linear portion of the first microphone signal using an adaptive filter.

4. The method of claim 1, further comprising, responsive to comparing the sum with the energy value of the audio captured at the second microphone, muting audio captured by the second microphone.

5. The method of claim 1, wherein the scaling factor is a positive number less than 1.

6. The method of claim 5, wherein the scaling factor is predetermined based, at least in part, on the relative distances of the first microphone and the second microphone from the loudspeaker.

7. The method of claim 5, wherein the first microphone and the second microphone are substantially non-distortive.

8. A memory storing instructions executable by at least one processor, the instructions comprising instructions to:

receive a first signal at an endpoint;

determine an energy value of the first signal;

emit audio at a loudspeaker, the audio based on the first signal;

collect audio at a first microphone, the audio including a first linear component corresponding to the first signal, and a first non-linear component corresponding to distortion of the first signal within the emitted audio;

emit, by the first microphone, a first microphone signal, the first microphone signal comprising a first linear portion corresponding to the first linear component of the collected audio and a non-linear portion corresponding to the first non-linear component of the collected audio;

determine an energy value associated with the non-linear portion of the first microphone signal;

transmit an energy signal to an echo canceller of a second microphone, the energy signal corresponding to the energy value of the non-linear portion of the first microphone signal multiplied by a scaling factor

capture audio at the second microphone, the captured audio including a second linear component corresponding to the first signal, and a second non-linear component corresponding to distortion of the first signal within the emitted audio, wherein the second linear component is attenuated relative the first linear component, and the second non-linear component is attenuated relative the first non-linear component;

determine an energy value of the audio captured at the second microphone;

receive the transmitted energy signal at the echo canceller;

determine, at the echo canceller, a sum of the energy value of the non-linear portion of the first microphone signal multiplied by the scaling factor with the energy value of the first signal;

determine, at the echo canceller, that the sum exceeds the energy value of the audio captured at the second microphone by a predetermined value; and

responsive to the determination that the sum exceeds the energy value of the audio captured at the second microphone by the predetermined value, allow transmission of the audio captured at the second microphone.

9. The memory of claim 8, wherein the instructions to determine the energy value associated with the non-linear portion of the first microphone signal comprise instructions to isolate the non-linear portion of the first microphone signal.

10. The memory of claim 9, wherein the instructions to isolate the non-linear portion of the first microphone signal comprise instructions to subtract the linear portion from the first microphone signal using an adaptive filter.

11. The memory of claim 8, wherein the instructions to transmit the energy signal to the echo canceller comprise instructions to transmit the energy signal to a double-talk detection unit of the echo canceller.

12. The memory of claim 8, wherein the scaling factor is a value between zero (0) and one.

13. The memory of claim 12, wherein the scaling factor is predetermined based, at least in part, on the relative distances of the first microphone and the second microphone from the loudspeaker.

14. The memory of claim 12, wherein the first microphone and the second microphone are substantially non-distortive.

15. A conferencing endpoint, the conferencing endpoint comprising:

at least one input, the input configured to receive a first signal, the first signal having an energy value;

at least one loudspeaker coupled to the input, the loudspeaker configured to emit audio, the audio based on the first signal;

at least one distortion detection module proximate the loudspeaker, the distortion detection module configured to collect audio, the collected audio including a first linear component corresponding to the first signal, and a first non-linear component corresponding to distortion of the first signal within the emitted audio, and further configured to emit a detection signal, the detection signal comprising a first linear portion corresponding to the first linear component of the collected audio and a non-linear portion corresponding to the first non-linear component of the collected audio;

at least one microphone configured to capture audio, the captured audio including a second linear component corresponding to the first signal, and a second non-linear component corresponding to distortion of the first signal within the captured audio, wherein the second linear component is attenuated relative the first linear component, and the second non-linear component is attenuated relative the first non-linear component;

at least one processing unit coupled to the input, the loudspeaker, the microphone, and the distortion detection module, the processing unit configured to;

determine an energy value associated with the non-linear portion of the detection signal;

apply a scaling factor to the energy value associated with the non-linear portion of the detection signal;

determine a sum of the scaled energy value of the non-linear portion of the detection signal with the energy value of the first signal;

compare the sum with an energy value of the captured audio; and

transmit the captured audio when the sum exceeds the energy value of the captured audio.

16. The conferencing endpoint of claim 15, wherein determining the energy value associated with the non-linear portion of the detection signal comprises isolating the non-linear portion of the detection signal.

17. The conferencing endpoint of claim 16, wherein isolating the non-linear portion comprises subtracting the linear portion from the detection signal using an adaptive filter.

18. The conferencing endpoint of claim 15, wherein emitting the detection signal comprises transmitting the energy detection signal to a double-talk detection module coupled to the processing unit.

19. The conferencing endpoint of claim 15, wherein the scaling factor is based, at least in part, on a gain of the microphone.

20. The conferencing endpoint of claim 15, wherein a distance from a central region of a top of the loudspeaker to the microphone is at least eighteen times greater than a distance from the central region of the top of the loudspeaker to the distortion detection module.

21. The conferencing endpoint of claim 20, wherein no portion of the distortion detection module is more than three millimeters distant from the portion of the loudspeaker to which the distortion detection module is closest.