A music structure inference algorithm based on symbolic data analysis

Gabriel Sargent (1) , Stanislaw Andrzej Raczynski (2) , Frédéric Bimbot (1) , Emmanuel Vincent (1) , Shigeki Sagayama (2)

1 METISS - Speech and sound data modeling and processing
2 UTokyo - The University of Tokyo

The present document describes a music structure inference algorithm submitted to the MIREX 2011 evaluation campaign (structural segmentation task). It consists of 3 stages : symbolic feature extraction, structural segment boundary estimation, and structural segment clustering. We consider as inputs chord estimations from the system of Ueda et al., expressed at the 2-beat scale. Beats and downbeats are estimated by the system of Davies et al. The structural segmentation step uses a regularity-constrained Viterbi approach. It assumes that the structure of pop songs is generally based on a few typical segments, whose sizes are called structural pulsation periods. The segments are then clustered according to their similarity, through the minimization of an adaptive model selection criterion.

Mots clés: Music structure; symbolic processing; Music information retrieval

Domaines: Multimédia [cs.MM]; Son [cs.SD]; Recherche d'information [cs.IR]

Gabriel Sargent, Stanislaw Andrzej Raczynski, Frédéric Bimbot, Emmanuel Vincent, Shigeki Sagayama. A music structure inference algorithm based on symbolic data analysis. MIREX - ISMIR 2011, Oct 2011, Miami, United States. ⟨hal-00618141⟩