Implement Real-Time Polyphonic Pitch Detection and Feedback System for the Melodic Instrument Player

摘要

This research proposes an automatic transcription-feedback system of music which help people to learn musical instruments by themselves. The focus of this research is piano. We develop real-time polyphonic pitch detection-feedback system. For 'polyphonic pitch detection', we use inner product based similarity measure with discriminant note detection threshold and top down attention. Also, we develop two parallel processes on simulink and matlab separately for real-time system. On simulink workspace, real-time recording and signal flow management is implemented. This system takes 2mins. 12secs. for analyzing 1min. piece and have accuracy of pitch detection as 79.33% for test case (Chopin Nocturne Op.9 N.2).