VoSA - Voice Separation Analyzer
The objective of the VoSA project is to develop a computer program to separate voices in polyphonic music using basic principles of music perception. The main applications of this research are:
- Music information retrieval: Most of the music we hear is polyphonic in nature, consisting of multiple voices sounding synchronously. Voice separation is necessary for matching melodic (single-voice) queries to polyphonic (multi-voice) databases, for example, in query-by-humming applications.
- Pitch spelling: accurate pitch spelling is dependent on contextual rules as well as melodic rules. A good voice separation algorithm will result in better implementation of voiceleading rules.