MIDI file format is popular for exchanging musical scores and ideas nowadays and massive of anonymous MIDI files are uploaded on the Internet. However, some of these MIDI files are edited with low music performance quality. This thesis researches the possibility for automatic enhancement of MIDI music performance quality, and we will start the investigation by editing MIDI piano tracks. We aim to accept universal MIDI files with various edit qualities; hence the enhancement algorithms will only be based on the basic component of MIDI, which is note pitches over time. After extraction of musical information from the limited musical data through various original algorithms, the music performance quality of the raw MIDI file would be enhanced by modification of dynamics and insertion of th...[ Read more ]

MIDI file format is popular for exchanging musical scores and ideas nowadays and massive of anonymous MIDI files are uploaded on the Internet. However, some of these MIDI files are edited with low music performance quality. This thesis researches the possibility for automatic enhancement of MIDI music performance quality, and we will start the investigation by editing MIDI piano tracks. We aim to accept universal MIDI files with various edit qualities; hence the enhancement algorithms will only be based on the basic component of MIDI, which is note pitches over time. After extraction of musical information from the limited musical data through various original algorithms, the music performance quality of the raw MIDI file would be enhanced by modification of dynamics and insertion of the sustain pedal. Besides, a GUI program named “MIDI Piano Enhancer” is developed for users to manipulate the automation process more easily. The results show significant MIDI music performance quality enhancement and suggest high feasibility of extension of the research to other instruments.