Structural Analysis of Musical Signals for Indexing, Segmentation and Thumbnailing

A musical piece typically has a repetitive structure. Analysis of this structure will be useful for indexing the digital music repository, segmenting music at transitions, and summarizing the thumbnails of music, all of which can be used in building efficient music retrieval and recommendation systems. This paper presents an algorithm that can automatically analyze the repetitive structure of musical signals. First, the algorithm detects the repetition of each segment of fixed length in a piece using dynamic programming. Second, the algorithm summarizes this repetition information and infers the structure based on heuristic rules. The performance of the approach is evaluated visually using figures for qualitative evaluation, and by two structural similarity measures for quantitative evaluation. The structural analysis result can be used in various ways in addition to the apparent use for indexing. The boundaries of the sections based on the structural analysis result have better consistency with transitions than boundaries produced by random segmentation. We can also employ different strategies based on structural analysis for music thumbnailing. Preliminary results obtained using a corpus of Beatles’ songs demonstrate the promise of these methods.