Bryan Pardo applies machine learning, probabilistic natural language processing, and database search techniques to auditory user interfaces for human-computer interaction. He takes a broader view of natural language than is traditional in computational linguistics, including timbre and prosody (timing, pitch contour, loudness), with an emphasis on music and speech prosody. Current projects include vocal interfaces and search engines for music databases, machine accompaniment of human musicians, and separation of sound sources in a stereo mix. New projects include cell phone-based karaoke games and parsing prosody in English. He has also been involved with other interesting projects, such as a tool to automatically create music videos and harmony recognition by machine. The corpus of musical examples he used for the harmony recognition project is available here.

If you are interested in creating working systems that understand music and/or speech prosody, then why not browse through some publications?