A Phase-Entrained Particle Filter for Audio-Locomotion Synchronization

Synchronized, audio feedback capitalizes on the innate human ability to entrain motion to a rhythmic, audio stimulus, and has been used in training athletes and treating gait disorders. For such feedback to be effective, it is necessary to close the loop to achieve mutual, bi-directional synchronization between human and machine. Although we know we can do the synchronization within an expensive, static motion capture facility, we seek low-cost, portable methods suitable for individual and outpatient use. This paper presents a phase-entrained particle filter that synchronizes to the periodic silhouettes of a gait, using only a simple camera as a sensor. Tests with eight participants show reliable synchronization, and timing errors in sonic stimulus generation less than 60ms, or a 1/32 note in a 120bpm march. While these errors are just audible, they do not confound mutual human-machine synchronization.