Bridging the Rhythmic Gap: A User-Centric Approach to Beat Tracking in Challenging Music Signals

In recent years, beat-tracking algorithms have exhibited considera­ble advancements in performance; however, challenges persist in the presence of signals featuring rhythmic dissonance. This paper introduces a user-centric approach for beat tracking in challenging music signals, adapting the current state-of-the-art through concise user-annotated regions. The proposed methodology demonstrates enhanced performance compared to existing techniques, particularly in demanding contexts with challenging rhythms. The human-in-the-loop strategy is scrutinized within the realm of computational ethnomusicology, addressing Music Information Retrieval (MIR) limitations in non-Western music. Evaluation on Colombian Bambuco, Uruguayan Candombe, and Steve Reich's Piano Phase, featuring polyrhythm, polymetre, and polytempo, respectively, assesses our method's efficacy. The paper culminates in a discussion of the findings within a wider context and a summary of potential future research avenues, underlining the importance of transfer learning and user-driven adaptation in advancing beat tracking for rhythmically diverse music collections.