Remote Monitoring of Tai Chi Balance Training Interventions in Older Adults Using Wearable Sensors and Machine...

Area of Science:

• Gerontology
• Biomedical Engineering
• Sports Science

Background:

• Tai Chi offers significant health benefits for older adults, including improved balance and reduced fall risk.
• Objective measurement of Tai Chi adherence and proficiency is challenging but crucial for optimizing health outcomes.
• Current methods for evaluating Tai Chi practice lack objectivity and scalability.

Purpose of the Study:

• To develop and validate a framework using wearable sensors and machine learning for monitoring Tai Chi training adherence and proficiency.
• To enable objective assessment of Tai Chi movement identification and performance quality.
• To provide a scalable solution for evaluating Tai Chi programs and informing clinical outcomes.

Main Methods:

• Utilized inertial measurement units (IMUs) to collect movement data from 32 participants performing six Tai Chi movements.
• Developed machine learning models, specifically Random Forest algorithms, for movement identification and proficiency assessment.
• Extracted features from IMU signals to train and evaluate the performance of the developed models.

Main Results:

• The movement identification model achieved a high accuracy of 90.05% (micro F1-score).
• Proficiency assessment models demonstrated high accuracy, with a mean micro F1-score of 78.64%.
• The framework successfully enabled detailed and objective analysis of Tai Chi movements.

Conclusions:

• Wearable sensors and machine learning provide a feasible and scalable method for analyzing Tai Chi practice.
• This approach can objectively enhance the evaluation of Tai Chi training adherence, learnability, and progression.
• The findings support the potential for optimizing Tai Chi program parameters to achieve better clinical outcomes.