Real Time Beat Tracking

Tracking beats while listening to music is an innate ability of humans. It is expressed in the form of tapping and clapping. Previously, there have been several non-causal attempts to mimic this action. Algorithmic beat tracking is a challenge for real-time implementation. We implement a causal beat tracking algorithm which includes onset detection, Beats Per Minute (BPM) estimation, beat phase estimation and beat prediction. In addition, we propose a novel cost function for beat prediction. The entire algorithm is implemented on Raspberry Pi 3 and creative light displays are used to highlight the underlying beat structure in music.