Virtual Humans and the Visual Perception of Human Movement

概要

論文の詳細を見る
We would like to create lifelike computer characters to populate our virtual worlds. These characters should be engaging and useful for various kinds of interactions between man and machine. One important requirement for these computer characters is that their movements be natural and realistic.Moreover, for many applications we would like our virtual humans to adapt rapidly to new situations and produce novel actions on the fly.The production of such natural and realistic movements in virtual humans is the primary goal of the ATR Cyberhuman project. The task of producing natural and realistic human movements involves problems in both the production and the perception of human movement. For this reason our work to create virtual humans involves combining expertise from the fields of computer graphics, biological motor control and visual perception. These areas of research have different and complementary perspectives on the same problem. For example, if we consider the basic problem of defining what is a "natural" or "realistic" movement we see that the perspectives of production and perception give us different insights. The perspective of motor control leads us to examine movements in terms of the physics of the bodies being moved and the control stategy used to plan and execute these movements. We can assume that if we had an adequate model of the physical world and a knowledge of the human motorcontrol strategy then we would be able to produce realistic and natural movements for our cyberhumans. However, to completely simulate all physical interactionns is overwhelmingly expensive in computation, and our current knowledge of human motor control is incomplete. Thus, engineering of virtual humans currently involves finding techniques for effective approximation of the physical world and the human motor controller. The perspective of visual perception considers how human movements are perceived and mentally represented. In this approach we must understand the constraints used to interpret the time-varying image of a movement and understand how these constraints shape our sensitivity to different aspects of movement. However, limited data exists to describe our sensitivity to human movement. This data suggests that we are able to distinguish between different styles of movement and, to a limited extent, the identity or emotional state of the actor performing this movement. But, little or no information exists on the cues used to perform these identifications. At present, work in the ATR Cyberhuman project involves getting a vitual human to play tennis. Tennis requires whole body movements that are performed in a variety of styles. Development of effective methods of creating a virtual tennis player and exploration of the features used by novice and expert tennis players to recogniqe tennis serves are current areas of investigation. It is hoped that these complementary investigations of the production and perception of human movement will produce general results that can be applied to the creation of virtual humans in a wide range of applications.
社団法人電子情報通信学会の論文
1998-09-18

Virtual Humans and the Visual Perception of Human Movement

スポンサーリンク

概要

著者

関連論文

スポンサーリンク