Learning motion patterns by imitating others\' motions and aquiring them as motions of itself is getting to be able to recognize and generate the acquired motion. Besides, description and memory of acquired motions are thought to be a unit of symbolic description of the motion and provide base for linguisitic descriptions.

In these points of view, we have been working on "Dynamics-Based Information Processing" ,which applies a deterministic difference equation, and "Statistic Information Processing", which applies a statistic and stochastic model.

Virtual human models are essential tools for humanoid robotics, cognitive information processing, CG animation, and medical applications. These pages briefly introduce our research on dynamics computation, generation, and measurement of motions of human figures.