A Real-Time Eating Detection System for Capturing Eating Behaviors In-the-Wild (Preprint)

BACKGROUND This paper describes a semi-automated eating detection system that leverages Ecological Momentary Assessment (EMA) questions to capture contextual factors upon detecting when an individual is eating. Our validation study demonstrates the efficacy of the system by deploying it in-the-wild among college students. OBJECTIVE This study builds a semi-automated eating detection system that leverages Ecological Momentary Assessment (EMA) questions to capture contextual factors upon detecting when an individual is eating. It also demonstrates the efficacy of the system by deploying it in-the-wild among college students. METHODS The eating detection system was deployed among 28 college students at a US institution over a period of three weeks. The participants reported various contextual information through EMAs triggered when the eating detection system correctly detected a meal episode. The EMA questions were designed after conducting a survey study with 162 students from the same campus. Responses from EMAs were used to define exclusion criteria. RESULTS Among the total consumed meals, 90% of breakfast, 99% of lunch, and 98% of dinner episodes were detected by our novel eating detection system. The eating detection system showed a high accuracy by capturing 95.67% of the meals out of 1,259 meals consumed by the participants. The eating detection classifier shows a precision of 80%, recall of 96%, and F1 of 87%. We found that over 99% of the meals were consumed with distractions. Such eating behavior is considered “unhealthy” and can lead to overeating and uncontrolled weight gain. Significant portions of meals were consumed alone (54.09%) in dorm rooms or apartment housing (31.19%). Our participants self-reported 63% of their meals as healthy. Together, these results have implications for designing technologies to encourage healthy eating behavior. CONCLUSIONS The presented eating detection system is the first of its kind to leverage EMAs to capture eating context, which has significant implications for wellbeing research. We reflect on the contextual data that has been gathered by our system and discuss how these insights can be used to design individual-specific interventions.