Segmentation and timbre- and rhythm-similarity in Electronic Dance Music
Bruno Rocha, Niels Bogaards, Aline Honingh

Abstract:
This report describes the digital humanities project on music similarity. The project is a collaboration between the University of Amsterdam and audio software company Elephantcandy. The project’s aim was to investigate timbre and rhythm similarity and to develop an application that finds similar segments of music. In this report three models are described, one for structural segmentation, one for timbre similarity, and one for rhythm similarity of electronic dance music (EDM). The segmentation algorithm performs well on an EDM dataset as well as on a standard MIREX dataset. The timbre similarity algorithm has been tested in a pilot study and preliminary results are presented. Issues related to segmentation and similarity are discussed.

Keywords: music;  similarity;  segmentation;  timbre;  rhythm