Children Behavior Tracking and Personal Identification By Multiple Kinect Sensors

In this paper, we proposed a children behavior tracking and personal identification system based on multiple Kinect sensors. In our system, two Kinect sensors are used, one is set horizontal in front of the children, to get their frontal face information, and the other one is set slanted, to get the whole scene with few occlusions among the children. Face, clothe color and moving information are used for personal identification, and this identification results are integrated to our extended Markov Chain Monte Carlo (MCMC) Particle Filter for robust tracking. The motion trajectory of each particular person can be extracted. The effectiveness of our method is proved through the experiments conducted in a nursery school.