Demonstration of a Jaw-Mounted Inertial Sensor For Eating Behavior Microstructure Characterization

Abstract

Capturing the fine-grained mechanics of eating, including individual bites and chews, is critical for addressing dietary health but remains challenging to measure passively in real-world environments. We present a wearable system designed for the unobtrusive, objective measurement of these eating mechanics. Our system pairs a minimalist, jaw-mounted inertial sensor placed over the mandibular condyle with a commodity smartwatch to capture high-fidelity mastication and food delivery gestures. Evaluated in a 30-participant multi-site study, the system achieved a bite detection F1 score of 0.833 and a chew-counting Mean Absolute Percentage Error (MAPE) of 8.565%. In this demonstration, attendees will observe a live, real-time visualization of 6-axis IMU data streamed from the jaw sensor as the presenter performs ambulatory motions, vocalization, and active food consumption, highlighting the system's ability to isolate chewing signals from extraneous noise.