Analyses of Textile Pressure-map Sensor Data of a Stuffed Toy for Understanding Human Emotional Physical Contact

Abstract

In this study, we investigate a stuffed-toy robot capable of estimating a user's emotional state via physical contact; the research is oriented toward the development of bidirectional human-robot communications. Textile-type pressure-map sensors are embedded in the robot's head and back, to facilitate detection of the location and strength of various touches via time-sequence data. Based on the recognition of such touching patterns as stroking, patting, and hugging, we conducted a verification of the estimation of the user's emotions using extracted features, such as strength and duration. The results indicate that emotional arousal was reflected by the number and strength of the touches, while the emotional valence emerged through the factor related to the context of the interaction. Our study thereby represents a possibility of understanding the user via physical contact.