Abstract
Extracting personal characteristics from human movement is important for applications such as generating behaviors of CG actors. In this paper, we propose a new method for extracting personal characteristics from 3D human movement. We introduce the eigen action space to represent personal characteristics of human movement. First, we estimate an average action from the set of 3D pose parameters from different people. Then, we estimate the difference of the average action and each persons 3D pose parameters. The difference actions can be transformed into the eigen action using KL decomposition. Because the personal characteristics are represented as a point in eigen action space, we can easily calculate the similarity measure of actions from different people. Also new actions are synthesized from eigen action space. In this paper, we shows the experimental results using the real video sequences to show the effectiveness of the proposed method.
