A Novel Kinect-Based Algorithm for Inferring the Position of the Lower Body Joints Using Human Gait Pattern

摘要

The Microsoft Kinect has gained wide popularity in health-related applications due to its low-cost, portability and ability to detect and track multiple body joints. Nevertheless, studies showed that the Kinect manifested inaccuracy when tracking the joints of the lower body, especially in non-ideal tracking conditions. In this work, a new algorithm was proposed for rectifying this tracking problem. As a proof of concept, the proposed algorithm was implemented only for the ankle joint because its readings contained the noisiest return values when compared to other joints. The proposed algorithm was used for inferring the joint's position in the 3D space. We were able to infer the ankle joint's position using the subject's gait pattern as expressed by the lower joints' angles and geometric relationships between these joints.