AEC (Acoustic echo cancellation) on ReSpeaker_Core_v2.0

Products & Technology ReSpeaker
#1

Hi,

I see that AEC (Acoustic echo cancellation) on ReSpeaker core is supported by using the librespeaker library, is there a way where in I can control (enable/disable) the AEC functionality on the core. Is there any test procedure or test suite which i can follow to validate this functionality.

Thanks,
Murtuza

#2

http://respeaker.io/librespeaker_doc/ Why don’t you look at the documentation and ask questions?

#3

Probably, because there’s no example of how to use AEC except some blind promises that it’s performed automatically by VepAecBeamformingNode?

If you take a look at how it’s designed at Amazon, their DSP algorithms watch the output audio waves to figure out what exactly should be subtracted from the input audio stream when it comes back to the microphone array.

If it seems too obvious to you, please share a code snippet with us of how to work both with input and output audio streams in parallel via provided API to effectively apply AEC. Or you can describe how this particular AEC algorithm is implemented in the Alango VEP package. As I have doubts that Respeaker can magically understand characteristics of the output waves w/o direct access to the output stream from code.

P.S. What I can say for sure is that AEC doesn’t work by just using a VepAecBeamformingNode from docs. You can try playing any mp3 through the connected speaker while active listening via librespeaker with “enabled” AEC. Respeaker doesn’t hear you at all if the music is louder than your voice. It means that AEC doesn’t work the way how it’s shown in Alango demos. So TS question is correct and I would like to hear any comments from librespeaker developers on that.

1 Like