The research on rehabilitation evaluation algorithm for frozen shoulder based on kinect

Shoulder disorders, such as frozen shoulder, cause significant upper limb dysfunction, severely affecting patients’ quality of life. Existing rehabilitation systems are costly and offer limited interactivity. This paper proposes a DTW algorithm with a relaxed endpoint diamond constraint band and time derivative sequence matching to improve evaluation and training efficiency. Using Kinect 2.0 for data acquisition and joint angles and distances as matching features, the algorithm was tested on a custom dataset in a Unity3D evaluation environment. The algorithm achieved a recognition accuracy of $96.79 \%$, with an average execution time of 1.114 s. The results show that the proposed algorithm efficiently assesses shoulder rehabilitation and is appropriate for patient training.