Front-view gait parameterization using arm movement

Abstract

We demonstrate a novel way of characterizing gait based on front-view upper-body movement. In particular, we take temporal measurements of the x-y position of arms and elbows of four persons in video, walking towards the camera. We compute the modulus of the Fourier Transform of the x-y positions. Preliminary results show that the Fourier transform modulus of the left hand x position was able to classify novel gait video footage of three out of four subjects by a shortest Euclidean distance rule.