Analysis of Weight-Bearing Squat Posture based on Deep Learning and Sports Science

This study investigates the biomechanics of weighted squats and the development of an AI-based fitness coach system utilizing deep learning and sports science. Employing BlazePose and MediaPipe for pose estimation, the system analyzes squat posture by calculating key angles and identifying squat phases. An experiment with 12 experienced participants performing squats at various weights was conducted. Data analysis revealed that increased weight results in greater forward lean and deeper knee flexion. The AI fitness coach system offers real-time feedback, data visualization, and progress tracking, thereby enhancing training effectiveness and safety. The findings demonstrate the system's potential to improve squat technique and prevent injuries.