Estimation of Emotions by Wearable Biometric Sensors Under Daily Activities

To examine the possibility of emotion estimation under daily activities with wearable sensors of objective biomarkers, we developed a discriminant model of the kind of emotion using pulse rate variability, skin temperature, body motion, and conversation. During 10 weeks of the experimental period, 5 healthy subjects recorded strongly conscious emotions (happy, relaxed, sad, and angry) at every 30 min during working hours. Significant association with the type of emotion was detected for heart rate, the variance of the frequency component of pulse rate variability, body movement, skin temperature, and conversation. Stepwise canonical discriminant analysis using these variables revealed that the kind of emotion can be expanded in two dimensions reflecting the levels of arousal and valence, respectively.