This study explains the metrical hierarchical structure of music that deals with onsets of musical events and the distances between them. The metrical structure is essential for the temporal organization of notes and the perception of rhythm. The paper shows the suitability of inter onset interval histograms to cover major metrical characteristics such as dominant metrical levels, regularity and onset density of a musical piece. Furthermore, experiments on automatic music similarity retrieval demonstrate the usefulness of the inter onset interval histogram for real world applications. The user chooses one song of a music collection for which he wants to get similar songs. The retrieval system sorts all other songs of the music collection according to the similarity of metrical properties to the given song. The results support the suitability of the histogram feature to retrieve similar songs within a rhythmic music collection.

Index Terms: music, rhythm, metrical structure, inter onset interval histogram, music similarity retrieval