KinPT: AI Telerehabilitation Assistant

Discipline: CS

Advisor: Dr. Chen Chen

Globally, 2.4 billion people need physical therapy, yet consistent access remains out of reach due to cost, geography, and lack of guidance between clinic visits. Research shows that 30-70% of PT patients fail to complete their home exercises correctly, risking re-injury or ineffective recovery. KinPT addresses these concerns by turning an Android smartphone into a real-time physical therapy coach. Using the device's built-in camera and computer vision technology, KinPT analyzes a patient's exercise movements frame by frame, tracking joint angles across 33 body landmarks to deliver instant audio and visual form corrections, counting only correctly performed repetitions. The mobile app is built around accessibility, with a clean, uncluttered interface and large text to ensure it’s elderly-friendly. To respect patient privacy, all processing happens on-device; no video is uploaded or stored externally. The SD1 prototype achieved high pose detection accuracy and rep counting accuracy on a mid-range Android device. The SD2 MVP will involve a completely polished UI, expand the live exercise experience with audio feedback, have the structure to expand beyond our squat exercise and introduce exercise plans, and introduce a therapist dashboard for remote patient monitoring.

Back to Showcase