Sequential Complexity as a Descriptor for Musical Similarity
This work addresses the challenge of capturing musically relevant temporal information for similarity and year prediction in Western popular music, representing an incremental improvement over existing methods.
The paper tackles the problem of determining musical similarity by proposing string compressibility as a descriptor for temporal structure in audio, achieving performance gains of 31.1% for similarity rating prediction and 10.9% for song year prediction when combined with bag-of-features descriptors.
We propose string compressibility as a descriptor of temporal structure in audio, for the purpose of determining musical similarity. Our descriptors are based on computing track-wise compression rates of quantised audio features, using multiple temporal resolutions and quantisation granularities. To verify that our descriptors capture musically relevant information, we incorporate our descriptors into similarity rating prediction and song year prediction tasks. We base our evaluation on a dataset of 15500 track excerpts of Western popular music, for which we obtain 7800 web-sourced pairwise similarity ratings. To assess the agreement among similarity ratings, we perform an evaluation under controlled conditions, obtaining a rank correlation of 0.33 between intersected sets of ratings. Combined with bag-of-features descriptors, we obtain performance gains of 31.1% and 10.9% for similarity rating prediction and song year prediction. For both tasks, analysis of selected descriptors reveals that representing features at multiple time scales benefits prediction accuracy.