A Biomechanics Motion Analysis System for Boxing Technique Evaluation using Computer Vision

Abstract

This paper presents a computer vision-powered boxing technique analysis system designed to assist athletes in improving their boxing form. The proposed system utilizes the MediaPipe pose estimation framework to extract upper-body skeletal landmarks and analyze boxing movements in real-time. By computing precise joint angles and body rotation metrics, the system compares user techniques against pre-computed professional reference poses for common boxing punches (jab, cross, hook, uppercut). We propose a rule-based system that calculates joint-specific and overall technique accuracies, providing actionable, detailed feedback on form and alignment. The system features a comprehensive frontend interface supporting real-time camera analysis and side-by-side video comparisons, incorporating robust performance optimizations for memory management and frame processing.