A Practical Approach for Recognizing Eating Moments with Wrist-Mounted Inertial Sensing

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Health Informatics

Background:

• Automated food intake monitoring is crucial for health research but faces challenges in practical, everyday application.
• Existing methods often require multiple sensors or specialized devices, limiting usability.

Purpose of the Study:

• To implement and evaluate a novel approach for inferring eating moments using readily available smartwatch technology.
• To assess the system's performance in real-world, free-living conditions.

Main Methods:

• Utilized 3-axis accelerometry data from an off-the-shelf smartwatch.
• Trained the system using data from 20 subjects in a semi-controlled laboratory setting.
• Evaluated performance in two free-living studies with varying durations.

Main Results:

• Achieved F-scores of 76.1% in one study (66.7% Precision, 88.8% Recall) and 71.3% in another (65.2% Precision, 78.6% Recall).
• Demonstrated successful recognition of eating moments in unsupervised, daily life scenarios.

Conclusions:

• This research contributes a practical, automated solution for everyday food intake monitoring.
• The developed system has potential applications in health research, personalized nutrition, and digital food journaling.