Classification of Dance Music by Periodicity Patterns

摘要: This paper addresses the genre classification problem for a specific subset of music, standard and Latin ballroom dance using method based only on timing information. We compare two methods extracting periodicities from audio recordings in order to find metrical hierarchy patterns by which style music can be recognised: first performs onset detection clustering inter-onset intervals; second uses autocorrelation amplitude envelopes band-limited versions signal as its periodicity detection. The relationships between are then used estimate tempo at beat measure levels hierarchy. interpreted musical note values, estimated tempo, meter distribution predict simple set rules. evaluated with test given CD cover, providing “ground truth” automatic measured.