Figure 1 shows the concept of virtual face-to-face communication system. A virtual actor (VA) is an interactive avatar which represents the talker's interactive behavior such as gesture, nodding, blinking, facial color and expressions, paralanguage, respiration and so forth on the basis of his or her verbal and nonverbal behavior in a virtual communication environment. Talkers can communicate through their VAs in the same virtual communication environment. Figure 2 shows an example of virtual face-to-face scenes from any viewpoint. The motions of head, both arms and body for each VA are represented on the basis of the positions and angles measured by 4 magnetic sensors (Polhemus FASTRAK) put on the talker's top of head, both elbows and the back of body. Two remote talkers can get embodied interaction awareness by observing the interaction of VAs including both his or her own VA and the partner's VA in the same environment. It is possible to analyze interaction by processing VAs' behavior such as cutting or delaying motion and voice.
While we have made the human VA representing human behavior precisely, we have made an abstract avatar of wave constructed by 6 × 6 cubes as shown in Figure 3 where the communication function of VA is simplified as a virtual wave (VW) by presenting only the head motion by one magnetic sensor put on