摘要: Understanding people’s dietary habits plays a crucial role in interventions that promote a healthy lifestyle. For this purpose, a multitude of studies explored automatic eating detection with various sensors. Despite progress over the years, most proposed approaches are not suitable for implementation on embedded devices. The purpose of this paper is to describe a method that uses a wristband configuration of sensors to continuously track wrist motion throughout the day and detect periods of eating automatically. The proposed method uses an energy-efficient approach for activation of a machine learning model, based on a specific trigger. The method was evaluated on data recorded from 10 subjects during free-living. The results showed a precision of 0.84 and a recall of 0.75. Additionally, our analysis shows that by using the trigger, the usage of the machine learning model can be reduced by 80%.