An Accelerometer based EyeGlass to Monitor Food Intake in Free-Living and Lab Environment

Abstract

This paper presents a smart eyeglass to monitor temporalis muscle movement for automatic food intake monitoring. The temple of the eyeglass is equipped with an accelerometer based sensing platform. The eyeglass is evaluated using four different classifiers for detection of chewing events during free-living studies. In addition, the in-lab studies are designed with two classifiers to detect chewing events and differentiate between various consumed foods based on their hardness. The system can achieve 86% accuracy, 82.14% precision, 85.49% recall and 82.23% F1-score for chewing detection in free living. For in lab studies, the system achieves 99.37% accuracy to detect chewing and 88% accuracy to differentiate between food based on hardness. The high accuracy results in both free living and in lab tests indicate that this eyeglass can be a preferable wearable to record food intake habits of people.