Melody and bass line estimation method using audio feature database

This paper proposes a musical instrument identification, melody and bass line estimation method of mixture sound signal for automatic music transcription. The method is based on sound feature verification using musical sound feature database such as tone (timbre) and pitch database. First, musical instrument is identified using MFCC. Then, melody and bass lines are estimated using time-frequency power spectrum peak positions of sound signal. Simulation results are given to show the effectiveness of our approach for musical instrument identification, melody and bass line estimation using mixture sound signals by MIDI.