A Holistic Eating Monitoring System Enabled by Textile Strain Sensors and Hierarchical Network

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Artificial Intelligence

Background:

• Continuous monitoring of physiological and biomechanical signals is crucial for understanding daily human dynamics.
• Existing wearable solutions lack integrated systems for fine-grained, continuous measurement of eating behavior.

Purpose of the Study:

• To develop a holistic system for continuous and detailed monitoring of eating behavior.
• To combine advanced sensor technology with deep learning for accurate analysis.
• To provide personalized feedback for healthy dietary guidance.

Main Methods:

• Utilized graphene-based textile strain sensors for comfortable, simultaneous acquisition of throat and jaw deformation signals.
• Implemented a hierarchical deep learning framework to model eating activities and biomechanical attributes from raw sensor data.
• Employed model visualization techniques for explainability and incorporated a large language model for personalized feedback.

Main Results:

• The system successfully enabled simultaneous and comfortable acquisition of physiological signals during eating.
• The hierarchical network effectively modeled eating-related activities and biomechanical attributes.
• Model visualization confirmed the correspondence between extracted data and physiological processes, enhancing explainability.

Conclusions:

• The proposed holistic sensing and learning system provides a foundation for personalized healthy dietary guidance.
• Enables dynamic monitoring essential for digestive health, metabolic regulation, and nutrition management.
• Offers a novel approach to understanding and improving eating behaviors through wearable technology and AI.