Human Body Motion Tracking using Color-Augmented Depth Sensing

Abstract

Motion tracking system using depth sensing camera has been widely used in motion game, sports, rehabilitation, and so on, which has limita- tions of self-occlusion, full body tracking, and person-occlusion in med- ical applications such as virtual rehabilitation. First, in order to under- stand limitations of conventional vision based method, we developed several rehabilitation and assessment programs using depth sensing cam- era. Although we confirmed the treatment effect and similarity with real assessment by the clinical testing, the occlusion and system integration problems could not cover a lot of movement of human. In this study, we examine this limitation in the developing process of several rehabilita- tion application, and resolve this problem by implementing a new track- ing method called Color Augmented Depth Sensing (CA-DT). The color based segmentation, Principle Component Analysis (PCA), and error weighted interpolation were used to develop CA-DT. Consequently, CA- DT system allowed the users to move freely without problems by partial occlusion, connection with other person. The result of comparison with a motion capture system of the passive marker type shows 0.97 correla- tion coefficient, which means that tracking ability of CA-DT is similar to conventional system. Furthermore, we considered patch size and shortened the armband type up to 5 cm breadth.