A practical approach for recognizing eating moments with wrist-mounted inertial sensing

摘要: Recognizing when eating activities take place is one of the key challenges in automated food intake monitoring. Despite progress over years, most proposed approaches have been largely impractical for everyday usage, requiring multiple on-body sensors or specialized devices such as neck collars swallow detection. In this paper, we describe implementation and evaluation an approach inferring moments based on 3-axis accelerometry collected with a popular off-the-shelf smartwatch. Trained data semi-controlled laboratory setting 20 subjects, our system recognized two free-living condition studies (7 participants, 1 day; participant, 31 days), F-scores 76.1% (66.7% Precision, 88.8% Recall), 71.3% (65.2% 78.6% Recall). This work represents contribution towards practical, monitoring, applicability areas ranging from health research journaling.