Real-time adapting captured human motions for tangible tele-meeting : Kinematic and dynamic approaches

Abstract

This paper describes two methods to adapt captured human motions to a humanoid robot for the tangible telemeeting system. The robot is used as a tangible avatar to imitate the movements of an actor, who is in the long distance. For adapting movements, the structural difference between the actor and the avatar is the main problem. The length of each arm segment of the actor differs from that of the avatar. Moreover, the number of joints and the degreeof-freedoms of each arm of the avatar are not the same as those of the actor. To resolve this problem, we introduce the kinematic and dynamic approaches. The two methods have been implemented and run in real-time at 50Hz. Experiments show that the methods work well adapting complex human arm movements. The comparison of two methods is also presented.