Contribution
Human-Guided Real-Time Beat Tracking Based on Predominant Local Pulse
* Presenting author
Abstract:
This work presents a real-time beat tracking system based on the Predominant Local Pulse (PLP) framework, extended with a human-in-the-loop control mechanism. While PLP provides a continuous estimate of the predominant pulse, its output can differ from a listener's perception or expectation depending on context, phrasing, or artistic interpretation. Such discrepancies may involve alternative metric levels such as double-time, half-time, or irregular subdivisions, as well as shifts in beat placement caused by syncopation or expressive timing. To address these perceptual differences, we propose a system that introduces two intuitive control parameters: a tempo-scaling parameter for adjusting the perceived metric level and a phase-offset parameter for refining beat alignment. Both parameters are integrated into the PLP computation through modifications of its local kernel structure. Real-time user input, such as tapping along with the music, enables the system to continuously adapt its internal pulse representation to perceptual and interpretive nuances. This human-guided approach enhances flexibility and expressiveness in interactive music systems while supporting both perceptual correction and creative rhythmic exploration. We evaluate its potential for improving beat perception, adaptability, and user engagement in real-time musical interaction.