Sounds like a simple app with mic input being sent to a yamnet-like audio classification model for a single target detection. Hardly anything innovative?