Trouble with Sound Source Separation

[dubai03nsr]

Hi,

I have a ReSpeaker Mic Array with 4 mics and am trying to use ODAS to perform sound source separation, but I haven't had much success. I've tried a variety of pairs of objects, including a drill vs a compressor, and two tone generator apps. I tried adjusting gain_sep from 1.0 to 10.0 but didn't observe any difference.

I also want to confirm that my method of obtaining the separated sound sources is correct. In my cfg file, under sss:separated:interface, I have type = "file"; path = "separated.raw";. After running odas.launch, I convert the separated.raw fie to a wav file (separated.wav) and then separate separated.wav into 4 wav files, each corresponding to a channel in separated.wav. However, the wav files don't contain isolated sound sources; i.e. both sound sources are audible in both wav files with audio (and the other two wav files are silent). Could you confirm that this is the proper way of obtaining isolated sound files?

[rnehrboss]

Hey @dubai03nsr
Were you ever successful with ODAS separtion on the respeaker 4 mic array?

[SeanTasker]

The output from sss needs to be processed with the data from sst. The channels from sss are dynamic. One source may use multiple channels over time depending on the overlap of sources in time.