A micro-macro thought experiment

It is pretty much universally acknowledged that we can quite accurately a person's mood from the tone of his/her voice. Could the sum of the tones of people's voices then tell us anything about the mood of an audience?

If, say, we have a few groups of audiences at different workshops and deliberately subject them to different discussions that made them happy, sad, angry, confused etc. And then measured the room noise with a microphone in one corner of the room. Is there a way we could reliably map noise patterns in discussions to actual moods without analyzing the actual contents of the discussions? Would a angry discussion 2-person conversation have a wave pattern as a room of vocal dissidents (in a non-trivial way, of course; similarity in volume because both are shouting matches doesn't count)? Can we use this to provide reliable feedback to the presenter?

Just something I thought about as I walked by what I believe was a BIOL126 prefect session just now.