Markerless Human Motion Tracking Using Microsoft Kinect SDK and Inverse Kinematics

摘要

Motion capture systems are used to gauge the kinematic features of the motion in numerous fields of research. Despite superb accuracy performance, the commercial systems are costly and difficult to use. To solve these issues, Kinect has been proposed as a low-priced markerless motion capture sensor, and its accuracy has been assessed using previous motion capture systems. However, in many of these studies, the anatomical joint angles captured using the Kinect are compared to the 3D rotation angles reported by the gold standard motion capture systems. These incompatibilities in the determination of the human joint angles can lead to higher error estimation. To accomplish a valid accuracy evaluation of the Kinect, we applied the inverse kinematics techniques in both Vicon and Kinect version 2 skeleton models to estimate lower body joint angles. The proposed method enabled us to capture the pelvic, hip, and knee joint angles using a single Kinect camera during gait. Moreover, the dependency of the proposed method to the position of the Kinect and the speed of the moving subject was investigated. In this study, the captured data of the Vicon motion capture system were used as ground-truth to assess the accuracy of the Kinect data. The results indicate the capability of Kinect in capturing human joint angles and also an affordable motion capture system applied in robotics and biomechanics applications.