Energy description of human motion capture data

In this paper, we present a novel energy-based model to describe human motion. According to biomechanics attribute of human body, we define kinetic and potential energy for each joint, upon which we convert the trajectory-based description of human motion into multi-dimensional energy series. As Dynamic Time Warping (DTW) distance is used to measure the similarity of motions, we employ an index for DTW, which is based on piecewise constant approximation, in order to speed up the query.