AI-Driven Smart Sportswear for Real-Time Fitness Monitoring Using Textile Strain Sensors.

Area of Science:

• Biomedical Engineering
• Sports Science
• Artificial Intelligence

Background:

• Wearable biosensors enable real-time human performance monitoring.
• Current systems struggle to non-invasively track breath-force coordination and muscle symmetry simultaneously.
• This limitation hinders applications in strength training and rehabilitation.

Purpose of the Study:

• To develop a wearable smart sportswear system for comprehensive exercise monitoring.
• To integrate graphene strain sensors and AI for real-time analysis of exercise execution.
• To address the limitations of existing biosensors in capturing complex biomechanical data.

Main Methods:

• Developed a wearable system with screen-printed graphene strain sensors and compact electronics.
• Implemented a deep learning framework (1D ResNet-18) for feature extraction and classification.
• Utilized tSNE and Grad-CAM for data visualization and model interpretability.

Main Results:

• Achieved 92.1% classification accuracy in identifying exercise execution quality across six conditions.
• Successfully distinguished between breathing irregularities and asymmetric muscle exertion.
• Demonstrated that the AI model captures biomechanically relevant features for robust interpretability.

Conclusions:

• The developed smart sportswear system provides a foundation for next-generation AI-powered athletic apparel.
• The system offers potential applications in fitness optimization, injury prevention, and adaptive rehabilitation.
• This technology enables seamless, non-invasive monitoring of critical performance parameters.