A Tai Chi balance training system for 0lder women based on a single-subject design: feasibility study of AI-sensor feedback

Abstract

Abstract: This study validates a real-time feedback system for Tai Chi training in elderly women using

wearable sensors and edge AI algorithms. The system integrates inertial measurement units (IMUs) and

surface electromyography (sEMG) to provide personalized feedback, improving balance and reducing fall

risks. The study employed a single-subject design with a 70-year-old healthy female participant, following

ethical guidelines to bypass complex IRB approval. The system achieved high accuracy in motion

recognition (92.3% precision) and real-time feedback (162±18 ms delay). It enhanced balance function,

with static stability improving by 55.8% and dynamic control by 35.4%. The system also optimized muscle

activation patterns, reducing injury risks. Despite limitations like sample size and environmental

adaptability, the study demonstrates the system’s feasibility and potential for community-based fall

prevention. Future work will focus on cross-gender validation, clinical integration, and community

deployment.