Development of an Artificial Intelligence System for Real-Time Analysis and Feedback in Silk Reeling Training of Chen-Style Taijiquan

Abstract

This paper introduces TaijiFlow AI, a web-based prototype serving as the core analytical module within the broader TaijiFlow Learning Ecosystem. Utilizing Artificial Intelligence and Computer Vision for real-time qualitative movement analysis, the system is designed to support independent practitioners of Chen-style Taijiquan by addressing the lack of expert feedback for foundational techniques such as Silk Reeling. By integrating Google MediaPipe Pose with an explainable, rule-based Heuristics Engine, TaijiFlow AI translates traditional Taiji principles into quantifiable kinematics. The system was developed under the ISO/IEC 29110 software engineering standard and ensures strict data privacy through 100% client-side processing on standard consumer hardware. Evaluation results show a 100% pass rate across all functional tests and high overall user satisfaction.