Motion tracking system for multi-user with multiple kinects

Abstract

User motion recognition through image recognition has attracted considerable attention with the development of the Kinect senor by Microsoft. The Kinect exhibits excellent motion recognition performance at a small expense via a motion tracking method utilizing an RGB-D camera. Further, the Kinect offers the advantage that applications can be easily written for it using the Kinect SDK. However, the Kinect also suffers from the disadvantage that it is unable to recognize an image “shielded” by an object because of the application of the vision recognition method by the camera. Consequently, it is difficult to obtain accurate data concerning user movement. Studies on observing an object from various directions with multiple Kinects are being conducted to compensate for this problem, and it is possible to acquire the movement data for only one user in these studies. This study implements the human body trace system for multiple users. We suggest a method to compensate for the location of the user and motion data that cannot be recognized by one Kinect using data acquired by multiple Kinects. © 2015 SERSC.