Many AEC algs seem to be linear or at least the non-NN algs seem to be. They can be extremely sensitive to timing even clock drift between input and output devices such as SpeexDSP.
I have wondered for a while if it would be possible to have a distributed ref signal by playing via a wireless audio system such as Snapcast https://github.com/badaix/snapcast
The AEC of any distributed wireless mic would have to cope with the possibility of multiple speakers of differing position and time domain and the tight sync Snapcast gives is by 'inching' the pcm to corect time sync.
I am presuming a Non linear AEC would cope with both these but never had anything to test and looked more into prioritising targetted speech extraction as a preference that doesn't have the restrictions of same device that linear AEC seems to imply.
Many AEC algs seem to be linear or at least the non-NN algs seem to be. They can be extremely sensitive to timing even clock drift between input and output devices such as SpeexDSP.
I have wondered for a while if it would be possible to have a distributed ref signal by playing via a wireless audio system such as Snapcast https://github.com/badaix/snapcast The AEC of any distributed wireless mic would have to cope with the possibility of multiple speakers of differing position and time domain and the tight sync Snapcast gives is by 'inching' the pcm to corect time sync. I am presuming a Non linear AEC would cope with both these but never had anything to test and looked more into prioritising targetted speech extraction as a preference that doesn't have the restrictions of same device that linear AEC seems to imply.