GUIDING ERGONOMIC HABITS WITH MACHINE LEARNING AND CAMERA-BASED ARTIFICIAL INTELLIGENCE FEEDBACK

Prolonged computers use fuels a growing epidemic of poor posture and related musculoskeletal issues, impacting quality of life and productivity. Addressing this, we propose a lightweight, real-time posture monitoring system designed for continuous background operation[1]. Utilizing Google's MediaPipe for pose detection and a heuristicbased scoring algorithm, our program analyzes key metrics like neck and torso angles [2]. The core challenge was objectively defining "good" vs. "bad" posture, which we addressed empirically with weighted metrics and an optimal threshold of 60.0. Experiments, using a 10,000-pose dataset, demonstrated 83.33% accuracy, with torso and neck angles proving most influential. This tool provides personalized end-of-day reports, leveraging AI (e.g., OpenAI's ChatCompletion API) to offer evidence-based recommendations[3]. Unlike specialized hardware or exercise-specific solutions, our camera-based application offers an accessible, continuous, and preventive approach for all prolonged computer users, fostering healthier digital habits.