A real time action scoring system for movement analysis and feedback in physical therapy using human pose estimation

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Computer Vision

Background:

• Human Pose Estimation (HPE) is crucial for physical therapy but faces challenges like motion blur, occlusions, and variable camera angles.
• These limitations hinder accurate movement assessment, especially in unsupervised home-based rehabilitation settings.

Purpose of the Study:

• To develop a novel action-scoring algorithm for enhanced physical therapy movement analysis.
• To improve the accuracy and robustness of automated rehabilitation monitoring systems.

Main Methods:

• The study integrates angular-based movement analysis with keypoint normalization techniques.
• Dynamic Time Warping (DTW) and Normalized Cross-Correlation (NCC) are used for movement comparison, with a fixed bounding box for tracking stability.
• A new angular calculation-based repetition counting mechanism is introduced to minimize angular noise.

Main Results:

• The proposed approach demonstrates high accuracy in joint angle measurements and repetition detection.
• The system shows increased robustness against occlusions and motion blur.
• Outperforms RepNet in accuracy and computational efficiency for video-based repetition counting.

Conclusions:

• The novel algorithm enhances the reliability of movement analysis in physical therapy.
• The system is suitable for real-time rehabilitation feedback, improving patient outcomes.
• This technology can expand access to automated physical therapy assessment tools.