When considered an improvement in communication between human and intelligent machine like computer and robot, it is natural that coordination between facial expression a s a nonverbal modality and voice as a verbal one may create more human-like communicative interaction. We have developed a construction of a mouth robot which can generate lip configurations for , I , U , e , and !bo! This paper describes 1) analysis of mouth shape formed on human face when helshe pronounces Japanese vowels, 2) determination of position of man-made muscles needed to form mouth shape, 3) small pneumatic actuator performances, 4) mechanical frame structure for installing man-made mouth skin, and 5) demonstration ofmouth shapes formed on the mouth robot for Japanese vowels. Japanese lla I1 11'11 I I I I I I I1
.INTRODUCTIONAs pointed out by Talceuchi and Nagao (1): Hara and Kobayashi (a), face-to-face conversation bctween humans is considered as a n ideal model in designing a n interface for human-machine communication since our daily communication between each other is fully based on the face-to-face communicat,ive interaction. The major feat,ure of face-toface communication is the multiplicity of communication channels. A channel is a communication medium associated with a particular encoding method, for instance; an auditory channel for carrying voices such as speech, utterance, intonation and so on; and the visual channel for face actions including nodding motion of a head, facial expressions and so forth. Thus the face-to-face communication is a multi-rnudal communicative interaction between humans and has been a long-term, hard work subject, in socio-psychology as well as in cognitive psychology (3).
Recently this style of communication h a s attracted a keen interest among researchers involved in anthropomorphic multi-modal human interfacefor human-machine communication (1) (2) (4).Takeuchi and Nagao (1) described that. in usual human face-to-face communication, many channels are used and different modalities are activated, and then the communication becomes highly flexible and robust,; so that failure of one channel i s recovered by other channels.Concerning the face-to-face communication: Mehrabian (5) indicated that only 7 % of message is d u e to linguistic language. 38 % is d u e to paralanguage. and 55 % of it is transferred by fqcia1 expressions. This implies that the facial expression is a major modality in human face-to-face communication. From t,his point, of view, there have been a quite few research works (1)-(3) on cognitive characteristics of human facial expressions as well as computer-graphic facial expressions. We have also studied the expressive characteristics of an artificial face robot, (6) wit,h respect to 6 t,ypical facial expressions.However when considered daily conversation as well as the data shown above, facial expression is found to be almost always integrated with voices. resulting in flexibilit,y and robustness in our message transfer in the face-t,o-face communication. We t,hus imagine that,, ...