Real-Time Human Body Posture Estimation Using a Stereo Vision Embedded System

This paper presents a real-time method to estimate upper human posture using a stereo vision embedded system. Because of using the stereo vision system that produces precise disparity image, it is possible to execute real-time posture estimation with high accuracy. Proposed method is as follows. By using image processing and trigonometry, the positions of head and hands were acquired and discriminated. Using these positions, shoulder positions are estimated. Finally, inverse kinematics is applied to estimate positions of elbows. The experimental result is presented to support validity of the real-time system and the algorithm.