Abstract
In this extended abstract, our work on the representation, indexing and retrieval of music data is summarized. We treat the rhythm, melody, and chords of a music object as music features and develop various data structures and algorithms to efficiently perform approximate and partial matching for the retrieval of music data [Liu99a], [Chen98], [Chou96]. In [Chen00a], we present the techniques for retrieving songs by music segments. A music segment consists of a segment type and the associated beat and pitch information. We also propose multi-feature index structures for exact and approximate searching on different features [Lee00].The problem of feature extraction is also studied. The repeating pattern is defined as a sequence of notes, which appears more than once in music objects. Choosing repeating patterns as the feature to represent the music objects meets both efficiency and semantic-richness requirements for content-based music data retrieval. We propose approaches to efficiently discover the repeating patterns of music objects in [Hsu98], [Liu99b]. We have also implemented Muse, a prototype system for content-based music data retrieval to illustrate the feasibility of the concepts we propose.