SR-POSE: A Novel Non-Contact Real-Time Rehabilitation Evaluation Method Using Lightweight Technology

Area of Science:

• Biomedical Engineering
• Computer Vision
• Rehabilitation Science

Background:

• Traditional rehabilitation movement assessment relies on cumbersome sensors or optical markers, limiting accessibility and increasing costs.
• There is a need for non-contact, accurate, and real-time methods for evaluating patient recovery and movement patterns.

Purpose of the Study:

• To develop and validate a non-contact, real-time system for accurate rehabilitation movement assessment.
• To provide a low-cost, user-friendly alternative to existing sensor-based assessment methods.

Main Methods:

• Utilized a lightweight pose detection algorithm, Sports Rehabilitation-Pose (SR-Pose), combined with a depth camera.
• Employed an E-Shufflenet network for feature extraction, a RLE-Decoder for key point regression, and a Weight Fusion Unit (WFU) for posture detection.
• Integrated 3D joint angle calculations and the Dynamic Time Warping (DTW) algorithm for sequence matching.

Main Results:

• Achieved an average detection speed of 14.32ms for human joint nodes.
• Demonstrated an average pose detection accuracy of 91.2%.
• Validated computational efficiency and effectiveness for practical rehabilitation applications.

Conclusions:

• The proposed non-contact approach offers a low-cost and user-friendly solution for rehabilitation movement assessment.
• SR-Pose system provides accurate, real-time analysis, enhancing the feasibility of remote and in-clinic patient monitoring.
• This technology holds significant promise for improving the accessibility and quality of rehabilitation services.