A 3-D Depth Information Based Human Motion Pose Tracking Algorithms

Abstract: Tracking human pose is the critical step of recognizing and analyzing the human motion on 3-D stereo vision, and it has great value and potential for applications of the machine vision. However, due to the complexity of human motion and background, most of the existing tracking methods for 3-D human motion gesture will add some extra restrictions on acquiring the human motion, and the tracking algorithms are often hard to obtain a robust algorithm performance. In this paper, a new human motion tracking method based on 3-D depth information is presented aiming to improve the tracking method quality. The algorithm makes a determination for the human body contour through analyzing the depth image information at first. Then the special skeletal points are estimated and tracked based on the 3-D depth vision images. Finally, the motion estimation is executed by using the three-step search algorithm, and the tracking pose is achieved naturally. Experimental results show the effectiveness of the algorithm, and it also has verified this method is feasibility and superiority on the aspect of acquiring human motion gestures and tracking compare to the similar method