Human walk pitch extraction by robot vision - towards human robot synchronized walking based on neural oscillator entrainment

Abstract

This paper presents a method to extract human walk pitch by robot vision. It is intended to be used for robot-human synchronized walking. Needless to say, in human-robot collaboration and interaction, synchronization between robot and human is crucial for safety, comfort, and sense of compatibility. It must be true even when a robot follows or accompanies its human master. In this work, the authors plan to realize the synchronization by utilizing entrainment of neural oscillator. A legged robot walks based upon self oscillation pattern generated by the oscillator without any stimulus. When the visually extracted human walk pitch data is fed into the neural oscillator as an external stimulus, the oscillator is entrained by the pitch data. In this paper the authors present both method and experiment to extract the walk pitch by tracking human heel. Experimental results support effectiveness of the method. Simulation on the entrainment of neural oscillator is also presented. In the simulation, experimentally extracted human walk pitch data is used as external stimulus. Simulation result demonstrates feasibility of the work.