I am a PhD. student in the analysis synthesis team at Ircam.
The title of my thesis is "L'extraction de structures d'une pièce sonore à partir de descripteurs audio", or "The extraction of structure of a musical piece starting from audio descriptors". The goal of the thesis is to create methods (in the form of software) to extract as many possible structure from musical piece: similarity, transformations, discontinuities on both micro- and macro-temporal scale.
More information (and non-information) can be found at my wikiwiki: Guakawaka.
Especially GuakaPhD, UsingPythonForScience and FreeMusic could be interesting to check out. And if you feel the slightest urge to add comments or something: don't hesitate to do so.