Project Title: Development of an AI System for Real-Time Analysis and Feedback in Silk Reeling Training of Chen-Style Taijiquan
Author: Mr. Weerayuth Uarjaipra
Degree: Master of Science in Software Engineering
Advisor: Asst.Prof. Parinya Suwansrikham, Ph.D.
Department / Institution: College of Art, Media and Technology (CAMT), Chiang Mai University
Abstract
Currently, individuals interested in practicing Chen-style Taijiquan face significant obstacles due to constraints in time, location, and a shortage of expert instructors, compelling many to rely on self-training through online media. However, practicing Silk Reeling (Chán Sī Jìng) independently presents a critical challenge: the lack of immediate qualitative feedback. Without this, practitioners remain unaware of biomechanical and movement errors as they occur. These issues led to the development of this independent study.
The independent study developed a simplified web application accessible through a browser, supporting standard webcam-equipped devices without requiring additional hardware. It integrates with MediaPipe Pose for real-time detection of 33 body landmarks, with all video frames processed locally on the client side to strictly preserve user privacy. The application supports four single-hand Silk Reeling patterns and is designed under a Body Generic architecture to accommodate practitioners of all ages and genders with diverse body types through three difficulty levels, ranging from Level 1 (Beginner) for the elderly to Level 3 (Advanced) for traditional training in a Bow Stance based on traditional Taijiquan principles.
Performance evaluation demonstrated that the system delivers smooth audiovisual feedback at an average of 45 frames per second (FPS) on standard computer hardware without a dedicated Graphics Processing Unit (GPU). User Acceptance Testing (UAT) with 15 participants indicated high user satisfaction (averaging 4.68 out of 5.00) and a 93.3% recommendation rate. Ultimately, this independent study illustrates the viability of applying transparent and Explainable AI (XAI) that guarantees user privacy to the preservation and transmission of the body of knowledge of Taijiquan in the digital age, establishing a robust prototype framework ready for future enhancement through Deep Learning techniques in subsequent phases.
Keywords: Chen-style Taijiquan, Silk Reeling, Heuristics, Pose Estimation, MediaPipe, Real-time Feedback, Web Application, ISO/IEC 29110, TaijiFlow, Learning Ecosystem