Segment Class Models for Music

par/by Darrell Conklin
Department of Computing
City University, London

In music, events can be grouped into higher level structures, for example notes into segments. This talk presents a representation for melodic segments with application to statistical modeling of music. Melody is represented as a sequence of segments, each segment being a sequence of notes. These segments are assigned to classes through a knowledge representation scheme which allows the flexible construction of abstract views of the music surface. Statistical models of segment class sequences are constructed from a corpus, and results on the task of musical style recognition will be presented. Possible extensions of the model will be proposed along with current research into other musical applications.