Human Pose Recognition Based on Skeleton Fusion from Multiple Kinects

Human pose recognition has attracted much attention in HCI, video games and so on. In this paper, a new method is proposed to recognize human pose based on skeleton fusion from multiple Kinects. Kinect can extract per-frame skeleton of human motion in real-time, which has problems of occlusions, data missing and errors. This paper adopts two Kinects to capture the human motion in different views to overcome these problems. Firstly, we use the skeleton tracking technology to obtain the 3D coordinates of 25 body joints; Secondly, we unify the joints coordinates which are captured by two Kinects in different views to a common world coordinates by using coordinates transform; Finally, we apply skeleton fusion algorithm to generate convinced human pose. From the experiment, we can easily find that the proposed method can effectively avoid the problems of occlusions, data missing and errors which often appear in single Kinect.