After 3 months of modelling, I've designed a first draft of a transcription algorithm (the "audio to bols" thing. .), whose error rate range from 1% to 10%. The system also learns parameters for an improved & more natural tabla sound synthesis. The program runs on Windows (java & C++)- a P300 MHz is required for realtime.
Unfortunately, I'm ruined and I'll have soon to find a job and give up these researches. I'm looking for funds to continue its development as a freeware (YOU can chose the platform : Windows, BeOS, QNX, MacOS, Linux, JavaVM), a job in the field of music software development or (musical) signal processing, or if possible start a thesis on related theoretical topics : robust estimation techniques, machine learning, minimum description length of sequences, statistical parsing, etc. .
I hope that some of you will be able to help me bringing this tool to the community of tabla lovers. .
