Participant controlled spatial AEC
View Patent ↗Methods and systems are provided for speaker separated spatial acoustic echo cancellation (AEC). A new approach is provided to combat the non-identifiability issue in multiple loudspeaker arrangements with shared echo cancellation processing. By associating echo channels with actual speakers (e.g., participants, talkers, users, individuals, etc.) rather than loudspeakers, the methods and techniques provided allow for continued use of mono AEC. Also provided is a straightforward speaker detection scheme for identifying different speakers in a stereo loudspeaker environment.
1. A method for acoustic echo cancellation comprising:
determining an active participant in an audio conference, wherein the active participant is one of a plurality of participants in the audio conference;
identifying a corresponding echo channel specific to the active participant;
converting at least one audio signal of the active participant received at a near-end of a communication path into a mono signal; and
performing echo cancellation on the converted mono signal based on the corresponding echo channel specific to the active participant,
wherein converting the at least one audio signal of the active participant into the mono signal includes selecting one of a plurality of loudspeaker signals rendered at the near-end of the communication path as the mono signal based on beam-forming.
2. The method of claim 1 , wherein the determining the active participant is performed at the near-end of the communication path.
3. The method of claim 1 , further comprising receiving the at least one audio signal of the active participant at the near-end of the communication path.
4. The method of claim 1 , further comprising:
responsive to determining the active participant in the audio conference, determining that no corresponding echo channel exists for the active participant; and
assigning an echo channel to the active participant, wherein the assigned echo channel is specific to the active participant.
5. The method of claim 2 , further comprising:
receiving, at the near-end of the communication path, participant information for the active participant; and
identifying the active participant based on the received participant information.
6. The method of claim 5 , wherein the received participant information is generated at a far-end of the communication path.
7. The method of claim 2 , further comprising:
identifying the active participant at a far-end of the communication path; and
receiving the identity of the active participant at the near-end of the communication path from the far-end.
8. The method of claim 1 , further comprising identifying the active participant using a plurality of signals at one end of the communication path.
9. The method of claim 8 , wherein the plurality of signals is a plurality of loudspeaker signals, and wherein the one end of the communication path is a near-end of the communication path.
10. The method of claim 8 , wherein the plurality of signals is a plurality of microphone signals, and wherein the one end of the communication path is a far-end of the communication path.
11. The method of claim 8 , further comprising:
obtaining pitch information for the plurality of signals; and
identifying the active participant based on the pitch information for the plurality of signals.
12. The method of claim 8 , further comprising:
deriving spatial information from the plurality of signals; and
identifying the active participant based on the spatial information.
13. The method of claim 8 , further comprising:
estimating direction of arrival of the plurality of signals; and
identifying the active participant based on the estimated direction of arrival of the plurality of signals.
14. The method of claim 13 , wherein the direction of arrival of the plurality of signals is estimated using a linear array structure.
15. The method of claim 8 , further comprising:
estimating time delay of the plurality of signals; and
identifying the active participant based on the estimated time delay of the plurality of signals.
16. The method of claim 8 , further comprising:
estimating a panning parameter for the plurality of signals; and
identifying the active participant based on the estimated panning parameter of the plurality of signals.
17. The method of claim 16 , wherein the panning parameter is estimated from a set of stereo audio signal data.
18. The method of claim 1 , wherein the corresponding echo channel for the active participant is obtained from a channel database.
19. The method of claim 4 , further comprising storing, in a channel database, the echo channel assigned to the active participant.
20. The method of claim 1 , wherein the loudspeaker signal is selected from the plurality of loudspeaker signals based on highest signal power.
21. The method of claim 1 , wherein converting the at least one audio signal into a mono signal includes:
determining an average of a plurality of loudspeaker signals rendered at the near-end of the communication path; and
using the average of the plurality of loudspeaker signals as the mono signal.
22. A method for acoustic echo cancellation comprising:
determining whether a participant in an audio conference has been identified;
responsive to determining that the participant has been identified, obtaining a corresponding echo channel specific to the participant;
responsive to determining that the participant has not been identified, indexing the participant and assigning an echo channel to the participant, wherein the assigned echo channel is specific to the participant;
selecting one of a plurality of audio signals rendered at a near-end of a communication path as a mono signal based on beam-forming; and
performing echo cancellation on the mono signal based on the corresponding echo channel specific to the participant or the assigned echo channel specific to the participant.
23. The method of claim 22 , wherein selecting one of the plurality of audio signals as the mono signal includes converting the one of the plurality of audio signals to a mono signal.
24. A method for acoustic echo cancellation comprising:
determining an active participant in an audio conference, wherein the active participant is one of a plurality of participants in the audio conference, and wherein the active participant is determined based on a panning parameter estimated from a set of stereo audio signal data at one end of a communication path;
identifying a corresponding echo channel specific to the active participant;
converting at least one audio signal of the active participant received at a near-end of the communication path into a mono signal; and
performing echo cancellation on the converted mono signal based on the corresponding echo channel specific to the active participant.