Dynamic Motion Tracking Based on Point Cloud Matching with Personalized Body Segmentation

Falling is a serious problem with the growing elderly population. In this sense, clinical institutions have implemented motor function assessment programs. In particular, the timed up and go test (TUG) is the most frequently applied clinical trial to assess the elderly walking ability in many clinical institutions and communities. In this study, we proposed a gait measurement system that can evaluate motor function in dynamic gait tests, such as the TUG test, using the point clouds of depth sensors (Kinect). The TUG test is a dynamic task that includes 3m of walking and turning motion. However, estimating joint positions using conventional methods that use Kinect skeleton function or point clouds is difficult. To solve these problems, before applying the iterative closest point algorithm, we proposed a method to move the segment model to a pre-estimated position and perform matching. In the accuracy verification experiments of several young people, the average error of each joint position was less than approximately 0.03 m, and the average error of the knee angle was approximately 4.54 to 5.13 degrees. These results indicate that the values estimated by the proposed method are useful as values for evaluating clinical tasks.