A head pose tracking system using RGB-D camera

Abstract. In this paper, a fast head pose tracking system is introduced. It uses iterative closest point algorithm to register a dense face tem-plate to depth data captured by Kinect. It can achieve 33fps process-ing speed without specific optimization. To improve tracking robustness, head movement prediction is applied. We propose a novel scheme that can train several simple predictors together, enhancing the overall pre-diction accuracy. Experimental results confirm its effectiveness for head movement prediction