pause ASR when speaker is outputting #173
Description
hello,
I am trying to build a voice-interactive system using deepspeech. everything is working properly in essence however I need the ASR to not recognize output of the system that is sent through the speakers. I can check if speaker is being used to output using os.popen('lsof /dev/snd/*).read() and checking if the speaker id is in the response.
However, I am not sure where to do this check to efficiently pause ASR. I have trying doing this check in vad_collector function, in the read/read_resampled functions, in the last part of the code where frames are added to the stream_context (model.createStream). Yet, I haven't been able to make it stop recognizing speaker output. If the speaker outputs 2 sentences back to back with a short pause in the middle, then I can get it to not recognize the first sentence but still it recognizes the second sentence.
Can anyone please help me out with ideas on how to pause ASR when speaker is outputting