Detection and suppression of returned audio at near-end
View Patent ↗Audio from a near-end that has been acoustically coupled at the far-end and returned to the near-end unit is detected and suppressed at the near-end of a conference. First and second energy outputs for separate bands are determined for the near-end audio being sent from the near-end unit and for the far-end audio being received at the near-end unit. The near-end unit compares the first and second energy outputs to one another for each of the bands over a time delay range and detects the return of the sent near-end audio in the received far-end audio based on the comparison. The comparison can use a cross-correlation to find an estimated time delay used for further analysis of the near and far-end energies. The near-end unit suppresses any detected return by muting or reducing what far-end audio is output at its loudspeaker.
1. A method of suppressing near-end audio acoustically coupled at a far-end and returned to the near-end in a conference, the method comprising:
determining first energy output of near-end audio sent from a near-end conferencing unit to a far-end conferencing unit by filtering the near-end audio into the first bands and determining first energy outputs of each of the first bands;
determining second energy output of far-end audio sent from the far-end conferencing unit to the near-end conferencing unit by filtering the far-end audio into the second bands and determining second energy outputs of each of the second bands;
detecting a return of the near-end audio in the far-end audio by
cross-correlating variations in the first and second energy outputs for each of the bands to one another at increments of time delay over a time delay range,
determining the time delay having a greatest cross-correlation for all of the bands,
estimating for each of the bands a ratio of the far-end energy to the near-end energy,
determining a result by multiplying the near-end energy by the ratio, and
determining whether the result is equal to the far-end energy,
wherein the detected return is caused by the near-end audio sent from the near-end conferencing unit to the far-end conferencing unit being acoustically coupled at the far-end conferencing unit and returned as part of the far-audio sent from the far-end conferencing unit; and
suppressing the detected return from output at the near-end conferencing unit.
2. The method of claim 1 , wherein the first and second bands each center at 400, 800, 1200, 1600, and 2000 Hz.
3. The method of claim 1 , wherein comparing the first and second energy outputs to one another over the time delay range comprises comparing the first and second energy outputs for each of the bands to one another over the time delay range.
4. The method of claim 1 , wherein detecting the return comprises:
offsetting the near-end energy versus time to match the far-end energy;
comparing the far-end energy outputs to the offset near-end energy outputs times the ratio for each of the bands; and
determining a count of how many of the bands have near-end energy outputs multiplied times the ratio that is at least less than or equal to an expected value.
5. The method of claim 4 , wherein detecting the return comprises declaring a presence of the return when the determined count is greater than a threshold.
6. The method of claim 1 , wherein the time delay range is from 0 to about 3-seconds.
7. The method of claim 1 , wherein determining the first and second energy outputs is done at predetermined sample times.
8. The method of claim 1 , wherein suppressing the detected return from output at the near-end conferencing unit comprises muting output of the far-end audio at a near-end loudspeaker.
9. The method of claim 1 , wherein suppressing the detected return at the near-end comprises reducing output of the far-end audio at a near-end loudspeaker.
10. The method of claim 1 , wherein the detected return is suppressed at the near-end conferencing unit for a predetermined time interval, until detection of the return ceases, or until speech having a magnitude greater than the return is detected in the far-end audio.
11. A non-transitory machine-readable medium having program instructions stored thereon for causing a programmable control device to perform a method of suppressing near-end audio acoustically coupled at a far-end and returned to the near-end in a conference, the method comprising:
determining first energy output of near-end audio sent from a near-end conferencing unit to a far-end conferencing unit by filtering the near-end audio into first bands and determining first energy outputs of each of the first bands;
determining second energy output of far-end audio sent from the far-end conferencing unit to the near-end conferencing unit by filtering the far-end audio into second bands and determining second energy outputs of each of the second bands;
detecting a return of the near-end audio in the far-end audio by
cross-correlating variations in the first and second energy outputs for each of the bands to one another at increments of time delay over a time delay range,
determining the time delay having a greatest cross-correlation for all of the bands,
estimating for each of the bands a ratio of the far-end energy to the near-end energy,
determining a result by multiplying the near-end energy by the ratio, and
determining whether the result is equal to the far-end energy,
wherein the detected return is caused by the near-end audio sent from the near-end conferencing unit to the far-end conferencing unit being acoustically coupled at the far-end conferencing unit and returned as part of the far-audio sent from the far-end conferencing unit; and
suppressing the detected return from output at the near-end conferencing unit.
12. A method of suppressing near-end audio acoustically coupled at a far-end and returned to the near-end in a conference, the method comprising:
sending near-end audio from a near-end conferencing unit to a far-end conferencing unit;
determining first energy outputs of first bands of the near-end audio at the near-end conferencing unit;
receiving far-end audio from the far-end conferencing unit at the near-end conferencing unit;
determining second energy outputs of each of second bands of the far-end audio received at the near-end conferencing unit;
detecting a return of the sent near-end audio in the received far-end audio by cross-correlation variations in the first and second energy outputs for each of the bands to one another at increments of time delay over a time delay range and determining the time delay having a greatest cross-correlation for all of the bands, the detected return being caused by the near-end audio sent from the near-end conferencing unit to the far-end conferencing unit being acoustically coupled at the far-end conferencing unit and returned as part of the far-end audio received at the near-end conferencing unit; and
suppressing the detected return from output at the near-end conferencing unit.
13. The method of claim 12 , wherein determining the first energy outputs of first bands of near-end audio comprises:
obtaining the near-end audio at the near-end conferencing unit;
filtering the near-end audio into the first bands at the near-end conferencing unit; and
determining the first energy outputs of each of the first bands.
14. The method of claim 12 , wherein determining the second energy outputs of second bands of far-end audio comprises:
obtaining the far-end audio received from the far-end conferencing unit at the near-end conferencing unit;
filtering the far-end audio into the second bands at the near-end conferencing unit; and
determining the second energy outputs of each of the second bands.
15. A conferencing unit, comprising:
a coder unit receiving near-end audio from a microphone and sending near-end audio signals to a far-end conferencing unit;
a decoder unit receiving a far-end audio signal from the far-end conferencing unit and sending far-end audio for output at a loudspeaker; and
a processor unit operatively coupled to the coder and decoder units and configured to:
filter the near-end audio into first bands,
determine first energy outputs of each of the first bands,
filter the far-end audio into second bands,
determine second energy outputs of each of the second bands, detect return caused by the near-end audio sent from the near-end conferencing unit to the far-end conferencing unit being acoustically coupled at the far-end conferencing unit and returned within the far-end audio received at the near-end conferencing unit, wherein to detect the return, the processor is configured to—cross-correlate variations in the first and second energy outputs to one another for each of the bands at increments of time delay over a time delay range, and determine the time delay having a greatest cross-correlation for all of the bands, and
suppress sending the detected return to a near-end loudspeaker.
16. The unit of claim 15 , wherein the conferencing unit is a near-end conferencing unit communicatively coupled to the far-end conferencing network via a network.
17. The unit of claim 15 , wherein the conferencing unit is a multipoint conferencing bridge unit communicatively coupled to a near-conferencing unit and the far-end conferencing network via a network.
18. The unit of claim 15 , wherein to detect the return, the processor is configured to:
estimate for each of the bands a ratio of the far-end energy to the near-end energy;
determine a result by multiplying the near-energy by the ratio; and
determine whether the result is equal to the far-end energy.
19. The unit of claim 18 , wherein to detect the return, the processor is configured to:
offset the near-end energy versus time to match the far-end energy;
compare the far-end energy outputs to the offset near-end energy outputs times the ratio for each of the bands; and
determine a count of how many of the bands have near-end energy outputs multiplied times the ratio that is at least less than or equal to an expected value.
20. The unit of claim 19 , wherein to detect the return, the processor is configured to declare a presence of the return when the determined count is greater than a threshold.
21. The unit of claim 15 , wherein suppressing the detected return from output at the near-end conferencing unit comprises:
muting output of the far-end audio at a near-end loudspeaker; or
reducing output of the far-end audio at the near-end loudspeaker.
22. The unit of claim 15 , wherein the detected return is suppressed at the near-end conferencing unit for a predetermined time interval, until detection of the return ceases, or until speech having a magnitude greater than the return is detected in the far-end audio.
23. The method of claim 12 , wherein detecting the return comprises:
estimating for each of the bands a ratio of the far-end energy to the near-end energy;
determining a result by multiplying the near-energy by the ratio; and
determining whether the result is equal to the far-end energy.
24. The method of claim 23 , wherein detecting the return comprises:
offsetting the near-end energy versus time to match the far-end energy;
comparing the far-end energy outputs to the offset near-end energy outputs times the ratio for each of the bands; and
determining a count of how many of the bands have near-end energy outputs multiplied times the ratio that is at least less than or equal to an expected value.
25. The method of claim 24 , wherein detecting the return comprises declaring a presence of the return when the determined count is greater than a threshold.
26. The method of claim 12 , wherein suppressing the detected return from output at the near-end conferencing unit comprises:
muting output of the far-end audio at a near-end loudspeaker; or
reducing output of the far-end audio at the near-end loudspeaker.
27. The method of claim 12 , wherein the detected return is suppressed at the near-end conferencing unit for a predetermined time interval, until detection of the return ceases, or until speech having a magnitude greater than the return is detected in the far-end audio.
28. The medium of claim 11 , wherein comparing the first and second energy outputs to one another over the time delay range comprises comparing the first and second energy outputs for each of the bands to one another over the time delay range.
29. The medium of claim 11 , wherein detecting the return comprises:
offsetting the near-end energy versus time to match the far-end energy;
comparing the far-end energy outputs to the offset near-end energy outputs times the ratio for each of the bands; and
determining a count of how many of the bands have near-end energy outputs multiplied times the ratio that is at least less than or equal to an expected value.
30. The medium of claim 29 , wherein detecting the return comprises declaring a presence of the return when the determined count is greater than a threshold.
31. The medium of claim 11 , wherein suppressing the detected return from output at the near-end conferencing unit comprises:
muting output of the far-end audio at a near-end loudspeaker; or
reducing output of the far-end audio at the near-end loudspeaker.
32. The medium of claim 11 , wherein the detected return is suppressed at the near-end conferencing unit for a predetermined time interval, until detection of the return ceases, or until speech having a magnitude greater than the return is detected in the far-end audio.