cs4 has the ability to analyse audio clips and convert the information into embedded meta-data text
which one would assume you could then do anything you wanted with.
the only problem with the system is that its inaccurate, like.. 80% on a good day with a professionally recorded voiceover track inaccurate.
i dont like your chances of pulling a decent result from a track with vocals and music, but you could try 🙂
~Peter Berthet
Sydney, Australia