Photorealistic speech agents and facially expressive robots

In an ideal human-computer interface, the computer would be able to understand and use modalities such as speech, hearing, facial gestures and body language in as realistic and natural a manner as a human. This paper discusses two kinds of computer-to-human interaction techniques which involve computerizea facial gestures: the first is the graphics representation of the human face projected on a video display; the second is the physical embodiment of facial expressions using an android robot. These simulated facial gestures may be accompanied by computerized speech synthesis, voice recognition capability or machine vision for recognition of the facial expressions of the human viewer. The author reviews both major techniques, then discusses current and proposed applications of this methodology.