Rhythm and periodicity detection in polyphonic music

We describe a novel approach for detecting perfect and imperfect periodicities in polyphonic music. The approach relies on beat and rhythm information extracted from the raw data after low-pass filtering. The beat and rhythm information is analyzed with a binary tree or trellis tree parsing depending on the length of the pauses in the underlying signal. This analysis yields accurate periodicity patterns at macro and micro scales. We illustrate the effectiveness of our approach using music segments from various cultures and explain its use in music classification and content-based retrieval.