Multimodal Evaluation of Stress Level Affected by Beverage Intakes

抄録

We propose an automated image analysis of facial expressions for measuring interventional effect of diet with respect to the mental aspect. Facial expression analysis that is noninvasive and noncontact with human subjects is promising in the sense that it can measure our natural behaviors due to emotional and biological states safely with low cost for lengthy period of time. This paper shows the analyses on the comparison between before and after the subjects ingest test samples, and also between effects of different test samples with respect to both the outputs from conventional sensors and image analyses of facial muscle activities. The subjects in our experiment (4 young adults, 2 males and 2 females) participated in a randomized and crossed over designed study for two separated experiment days. Water and green tea (250ml, respectively) were used as test samples. The measurements were recorded 1) before the ingestion, 2) 30 minutes after the ingestion, and 3) 60 minutes after the ingestion. During the time between each measurement, the subjects were asked to take an arithmetic mental task (AT) for 20 minutes as a stressor. Our results suggest a possibility that automated image analysis of facial expressions is one of the useful tools for evaluating the changes on mood and stress level affected by foods.