Development of Real-time Compensatory Movement Detection System in Neck Stretching Exercises for Office Syndrome Using Pose Estimation Technology

Abstract

The objectives of this research were 1) to develop a prototype system for verifying correctness and counting repetitions of neck stretching exercises to minimize compensatory movements in Office Syndrome patients, 2) to evaluate the system's accuracy and performance, and 3) to assess user satisfaction. The developed system operates on a personal computer using a standard webcam. It utilizes Python programming language and the MediaPipe Framework to detect pose landmarks. Geometric analysis algorithms were implemented to calculate neck and shoulder angles. The system was specifically designed to detect "Shoulder Elevation," a common compensatory movement, by setting a shoulder angle threshold at 8 degrees and a neck flexion threshold at 25 degrees. Real-time visual feedback is provided to alert users of incorrect posture, and the repetition counter increments only when the exercise is performed correctly. The sample group for performance testing and satisfaction assessment consisted of 30 students and personnel. The experimental results, based on 1,200 trials compared with expert evaluation, indicated an overall accuracy of 96.67%. The system demonstrated a sensitivity of 97.00% in detecting cheating postures and a specificity of 96.33%. User satisfaction assessment revealed a "very high" satisfaction level ( = 4.62, S.D. = 0.50), particularly regarding the effectiveness of real-time alerts and confidence in exercise correctness. This research demonstrates that low-cost pose estimation technology can be effectively applied as a self-rehabilitation tool.