T he fast pace in computing, graphics, and networking technologies plus the demands of real-life applications have impelled the development of more realistic virtual environments (VEs). Realism depends on a believable appearance and simulation of the virtual world and also implies a nat- ural representation of participants. This representation includes ■ a visual embodiment of the user, ■ a means of interacting with the world, and ■ a means of feeling various attributes of the world using the senses. Realism in participant represen- tation involves two elements: believ- able appearance and realistic movements. This becomes even more important in multiuser net- worked virtual environments (NVE), since the participants’ representa- tions are used for communication. We can define an NVE as a single environment shared by multiple participants connected from differ- ent hosts. The participants’ local program typically stores the whole or a subset of the scene description, and they use their own avatars to move around the scene. Rendering takes place from their own view- points. This avatar representation in NVEs has crucial functions in addi- tion to those of single-user virtual environments: ■ Perception (to see if anyone is around) ■ Localization (to see where the other person is) ■ Identification (to recognize the person) ■ Visualization of others’ interest focus (to see where the person’s attention is directed) ■ Visualization of others’ actions (to see what the other person is doing and what she means through gestures) ■ Social representation of self through decoration of the avatar (to know what the other participants’ task or status is) Using virtual human figures for avatar representation fulfills these functions realistically, providing a direct relationship between how we control our avatars in the virtual world and how our avatars move in response to this control. Even with limited sensor information, we can construct a virtual human frame that reflects the user’s activities in the virtual world. Slater and Usoh 1 indicated that using even a simple virtual body increas- es the sense of presence in the virtual world. (See the sidebar “Definitions and Concepts” for explanations of the terms used in this article.) NVEs with virtual humans are emerging from two threads of research with a bottom-up tendency. First, over the past several years, many NVE systems have been created using various types of network topologies and computer architectures. The practice is to bring together different, previously developed monolithic applications within one standard interface, building multiple logical or actual processes to handle separate elements of the VE. Second, at the same time, virtual human research has developed to the point of provid- ing realistic-looking virtual humans that can be ani- mated with believable behaviors in multiple levels of control. Inserting virtual humans in the NVE is a com- plex task. The main issues include ■ selecting a scalable architecture to combine these two complex systems, ■ modeling the virtual human with a believable appearance for interactive manipulation, ■ animating the virtual human with a minimal number of sensors to achieve maximal behavioral realism, and ■ investigating different methods to decrease the net- working requirements for exchanging complex vir- tual human information. Tolga K. Capin, Hansrudi Noser, and Daniel Thalmann Swiss Federal Institute of Technology Igor Sunday Pandzic and Nadia Magnenat Thalmann University of Geneva Virtual Human Representation and Communication in VLNet 0272-1716/97/$10.00 © 1997 IEEE 3D and Multimedia on the Information Superhighway 42 M arch-April 1997 Using virtual humans to represent participants promotes realism in networked VEs. Different message types used to animate the human body and face impose varying network requirements, as analyzed here. .