Learning optimal features for music transcription | IEEE Conference Publication | IEEE Xplore