Communication atmosphere in humans and robots interaction based on the concept of fuzzy atmosfield generated by emotional states of humans and robots

Abstract

Communication atmosphere based on emotional states of humans and robots is modeled by using Fuzzy Atmos-field (FA), where the human emotion is estimated from bimodal communication cues (i.e., speech and gesture) using weighted fusion and fuzzy logic, and the robot emotion is generated by emotional expression synthesis. It makes possible to quantitatively express overall af-fective expression of individuals, and helps to facilitate smooth communication in humans-robots interaction. Experiments in a household environment are performed by four humans and five eye robots, where emotion rec-ognition of humans based on bimodal cues achieves 84% accuracy in average, improved by about 10 % compared to that using only speech. Experimental results from the model of communication atmosphere based on the FA are evaluated by comparing with questionnaire surveys, from which the maximum error of 0.25 and the mini-mum correlation coefficient of 0.72 for three axes in the FA confirm the validity of the proposal. In ongoing work, an atmosphere representation system is being planned for casual communication between humans and robots, taking into account multiple emotional modalities such as speech, gesture, and music.