noise suppression with librespeaker in RespeakerCorev2

How do you think that if we capture audio via alsa api and redirect the processed audio stream with alsa-aloop? Then snips or other ASRs can capture the processed audio stream easily.