The Research On The Key Techniques Of The Intelligent Virtual Human's Action Control

Virtual Human is the expression of geometric and behavioral characters of human in the space generated by computer (virtual environment)^[1]. It is not only to simulate the appearance of the human, but also to have the real simulation of the action to exactly achieve the human's reappearance in the three-dimensional environment. But human's body has more than 200 freedom-degrees, which can complete the complex actions, and it is one objective of the Virtual Human's research to how to simulate the action.People have brought out many various methods of Virtual Human's action control from the early key- frame's method to the action control based kinetics. Although these methods have respective merit, no one is very perfect. The reason is that the human's action is not only a mechanical action, but also has the emotional factors in it. So this paper brings forward the new method which combines the Affective Computing,Artifical Neural Network and the Virtual Human, studies the Affective Computing, brings forward its model, makes Virtual Human has the emotions and the ability of study by itself. The paper introduces the work and evolvement in these works, the detailed content as follow:The action control of Virtual Human includes two aspects which are low and high action control. The low action control is very mature. This paper adopts the key-frame control method after reviewing various low control methods and integratings the virtual human's body modeling into the VRML technic. In the high action control, this paper emphasizes research on the route layout, based the shortest tangent arithmetic bring my tangent equation and improve the arithmetic.Affective Computing is going to make the computer having the ability of observing,understanding and creating the emotional character as the people, and ultimately to make the computer natural,friendly and vividly alternate like the people. The paper takes the affective computing into the intelligent Virtual Human to make its action more personalized and intelligent. Via estimating the human's Real-time inner state(including the emotional state), to have the important influence on the selection of human's action. Different inner state can lead different selection of actions.Artifical Neural Network is recently an important problem in the research of virtual human. This paper brings the Artifical Neural Network into the training of the virtual human, and after the training it can make the virtual human study by itself, do the simple logistic judgment, and make the action selection of virtual human intelligent.