Melody extraction on MIDI music files

Abstract

In this study, we propose a new approach to extract monophonic melody from MIDI files and provide a comparison of existing methods. Our approach is based on the elimination of MIDI channels those do not contain melodic information. First, MIDI channels are clustered depending on pitch histogram. Afterwards, a channel is selected from each cluster as representative and remaining channels and their notes are removed. Finally, skyline algorithm is applied on the modified MIDI set to ensure accuracy of monophonic melody. We evaluated our approach within a test bed of MIDI files, composed of variable music styles. Both our approach and the results from experiments are presented in detail.