Random idea: Using MIDI for WAV file compression
Since even mobile devices now have basic General MIDI support now, what about approximating digital audio as a stream of very short MIDI notes?
Encoding seems like it would be reasonably simple - do a best fit of each 100ms audio sample with a library of cached samples of a few hundred MIDI samples converted to raw WAV format, then stream the MIDI info. Obviously this could be optimized with some math smarts and statistical analysis on voice and music data. It wouldn't handle polyphonic music well, and would be pretty darn lo-fi, but it seems like a good way to compress audio since both sides of the communication have a common library of audio information.
Of course this is not so useful now that bandwidth is plentiful, but it still could make for at least an interesting BSCS project, right?
Usually any ideas I come up with are unworkable, have been patented the year before or are unpopular for some reason I don't understand. Which is this?