Eating Episode Detection with Jawbone-Mounted Inertial Sensing

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Human-Computer Interaction

Background:

• Automated Dietary Monitoring (ADM) traditionally uses necklace-mounted sensors for eating detection.
• Necklace-based systems face challenges with motion artifacts and sensor misalignment due to body movement.
• Existing methods require improvement for accuracy and user inconspicuousness.

Purpose of the Study:

• To develop and evaluate a novel wearable inertial sensing platform for precise eating detection.
• To overcome limitations of necklace-based sensors by directly mounting on the jawbone.
• To assess the system's performance in both laboratory and naturalistic environments.

Main Methods:

• Developed a small, wireless inertial sensing platform for direct jawbone mounting.
• Implemented a data analysis pipeline to identify eating episodes from inertial sensor data.
• Evaluated the system using leave-one-participant-out cross-validation (LOPO-CV) in controlled and real-world settings.

Main Results:

• Achieved 91.7% precision and 91.3% recall in laboratory settings (n=9).
• Demonstrated 92.3% precision and 89.0% recall in naturalistic settings (n=14).
• Showed a significant improvement (>10% in F1 score) compared to necklace-based approaches.

Conclusions:

• Directly mounting inertial sensors on the jawbone offers superior eating detection accuracy.
• The developed system is more inconspicuous and suitable for clinical adoption.
• This approach advances automated dietary monitoring with improved wearable technology.