Hello, anonymous submitter,
About I year ago I've started a small project with a friend, written in C++, that does some improvements to lossless audio compression (currently to FLAC). It calculates the autoconvolution of the audio signal to find similar parts in the audio. Especially for electronically generated music, this can be used to predict other parts in the audio, and thus reduce the entropy of the signal (i.e. compress the file better).
If you (or anybody else) are interested in further development, I would open-source the project, and may be start working on it again a bit myself. Just drop me a line, at polarspaceflo (at) googlemail (dot) com.