Resumen de A Review of Musical Rhythm Representation and (Dis)similarity in Symbolic and Audio Domains
Diogo Cocharro, Gilberto Bernardes, Gonçalo Bernardo, Cláudio Lemos Fonteles
In this paper, we review computational methods for the representation and similarity computation of musical rhythms in both symbolic and sub-symbolic (e.g., audio) domains. Both tasks are fundamental to multiple application scenarios from indexing, browsing, and retrieving music, namely navigating musical archives at scale. Stemming from the literature review, we identified three main rhythmic representations: string (sequence of alpha-numeric symbols to denote the temporal organization of events), geometric (spatio-temporal pictorial representation of events), and feature lists (transformation of audio into a temporal series of features or descriptors), and twofold categories of feature- and transformation-based distance metrics for similarity computation. Furthermore, we address the gap between explicit (symbolic) and implicit (musical audio) rhythmic representations stressing that a greater interaction across modalities would promote a holistic view of the temporal music phenomena. We conclude the article by unveiling avenues for future work on (1) hierarchical, (2) multi-attribute and (3) rhythmic layering models grounded in methodologies across disciplines, such as perception, cognition, mathematics, signal processing, and music.
