Sequential rhythmic information retrieval for audio similarity matching

Often, two music pieces may be rhythmically similar, despite being composed of different instrument sets. These instrument sets though acoustically different, play similar rhythmic roles and hence project similar rhythmic feel. Also, the accompaniment music, besides the dominant drums or brass, may influence one´s perception of the similarity/dissimilarity of two music pieces. These motivate our study on the rhythmic structure patterns of music compositions at different salience levels. In this paper, we illustrate a sequential rhythmic information retrieval approach, based on the salience of detected onsets. Primary experimental findings on a database of R&B and Hip-hop music have shown the effectiveness of the proposed rhythmic information retrieval method for audio similarity matching.