Real world video avatar: transmission and presentation of human figure

Video avatar (Ogi et al., 2001) is one methodology of interaction with people at a remote location. By using such video-based real-time human figures, participants can interact using nonverbal information such as gestures and eye contact. In traditional video avatar interaction, however, participants can interact only in "virtual" space. We have proposed the concept of a "real-world video avatar", that is, the concept of video avatar presentation in "real" space. One requirement of such a system is that the presented figure must be viewable from various directions, similarly to a real human. In this paper such a view is called "multiview". By presenting a real-time human figure with "multiview", many participants can interact with the figure from all directions, similarly to interaction in the real world. A system that supports "multiview" was proposed by Endo et al. (2000), however, this system cannot show real-time images. We have developed a display system which supports "multiview" (Maeda et al., 2002). In this paper, we discuss the evaluation of real-time presentation using the display system.