Automatic Music Transcription Using Genetic Algorithms and Electronic Synthesis David Lu ! ! April

Polyphonic music transcription is a computationally difficult problem that is still widely regarded as unsolved. This paper describes one method for the problem of pitch recognition with multiple instruments. A genetic algorithm produces generations of possible transcriptions that are evaluated for fitness by comparing the transcription’s generated audio to the original sound file. By sharing and modifying the genetic material of thousand of individual possible transcriptions, ultimately, the algorithm arrives at an accurate transcription for the original audio, or at least something that produces a similar wave form.