Real-time 3D Motion Capture
- Alex Pentland ,
- Christopher Wren ,
- David Harwood ,
- Ismail Haritaoglu ,
- Larry Davis ,
- Thanarat Horprasert
We describe a real-time 3D computer vision system for detecting and tracking human movement and providing a person with control over the movement of a virtual puppet. Multiple cameras observe a person; silhouette analysis and template matching achieve real-time 3D estimation of human posture. A dynamics/kinematics model of human body and Kalman filter are utilized to help the tracking process as well as to interpolate some joint locations. All estimation and rendering processes run in real-time on a PC based system.