Currently the expression measurement API supports speaker-based diarization (identify_speaker param) only. For use cases where speakers are diarized by channels, we'd have to send each audio channel separately to the API for processing.

This requirement is for supporting channel-based diarization.