Automated Assessment of Sit-to-Stand Test Using Machine Learning and Image Processing

The Sit-to-Stand (STS) test is a common measure of lower limb strength and functionality, particularly in rehabilitation and clinical settings. This paper presents a novel automated system for assessing STS performance using machine learning-based pose detection. By integrating the MediaPipe pose estimation library into a web-based interface, we track key joint angles in real-time to count repetitions and evaluate user performance. The system leverages the angles formed between the shoulder, hip, and knee joints to detect sit and stand transitions, with thresholds set for sitting ($\lt 120$ degrees) and standing ($\gt 160$ degrees). The system accurately tracks STS repetitions over a 60second period, providing an automated count and analysis of the user’s performance. Initial testing with multiple users showed high accuracy when facing specific directions, although further optimization is needed for right-side detection. Future work will focus on improving system robustness for users with balance issues and in clinical populations. This automated system offers a scalable solution for remote health assessments and real-time feedback during physical rehabilitation.