The "Smart Dining Table": Automatic Behavioral Tracking of a Meal with a Multi-Touch-Computer.

Area of Science:

• Behavioral Science
• Human-Computer Interaction
• Biomedical Engineering

Background:

• Understanding human eating behavior is crucial for regulating food intake and preventing obesity and eating disorders.
• Current methods for mealtime behavioral evaluation lack automatic data collection and online analysis capabilities.
• There is a need for interactive technologies to enhance the study of eating behaviors at the meal table.

Purpose of the Study:

• To examine the feasibility of a novel technology for automatic detection and classification of bites during laboratory meals.
• To evaluate the accuracy of a SUR40 multi-touch tabletop system combined with a Kinect camera for bite detection.
• To assess an alternative bite detection method using a Myo armband as a more sensitive approach.

Main Methods:

• A SUR40 multi-touch tabletop with infrared camera tracked tagged plates for bite detection.
• A Kinect camera recorded meals for verification and provided gesture detection.
• A second experiment utilized a Myo armband with a nine-axis accelerometer for bite detection, analyzing magnetometer data.

Main Results:

• The initial SUR40 system achieved 67.5% sensitivity and 82.4% classification accuracy for bite detection.
• The Myo armband demonstrated improved sensitivity (86.1%) compared to the Kinect (60.5%).
• The Myo armband also showed higher precision (72.1%) than the Kinect (42.8%) in positive predictive value.

Conclusions:

• The SUR40 tabletop combined with the Myo armband is a feasible technology for automatic bite detection and classification.
• This combined system offers adequate accuracy for various applications in studying eating behaviors.
• The findings suggest potential for developing advanced tools for healthy eating interventions and research.