Emotion estimation by acceleration pulse wave analysis

抄録

The prevalence of mood disorders is increasing worldwide, making mental health self-care necessary. One possible solution to address this need is the use of wearable technology that monitors and records moods. In this study, we attempted to (1) estimate the emotional states of joy, anger, and neutral from blood volume pulse intervals and (2) evaluate the potential applicability of blood volume pulse-based mood mapping in wearable devices. The pulse wave data available on the Massachusetts Institute of Technology (MIT) Affective Computing Group website were used. Acceleration pulse waves were analyzed using second-order differential calculus. Mean NN interval, standard deviation of NN intervals, and coefficient of variation of RR intervals were analyzed as emotional features, based on which a three-state classification model was created via linear discriminant analysis. The classification accuracy at the pulse wave measurement time of 30 seconds was 57%, whereas that at 100 seconds was 53%. Our findings indicate that mood estimation using acceleration pulse wave analysis has potential application in wearable technology for mental health self-care and warrant further research to strengthen the data.