Recent advances in motion capture makes it possible to construct natural motions in real time. In motion capture method, rotations of each joints are captured from the real human's motions and stored in the computer, and these data are fed to the virtual agent or human model to animate their body motions. Because motion captured data doesn't accomplish general goal position, motion variation technique that edits motion captured data is needed in order to form goal-directed motion. The ultimate objective of this thesis is to develop a method to generate the realistic human motion when the goal of the body's control point is specified. The main difference between our objective and general motion capture technique, is that we are interested in goal-directed motion while others just try to show the motions without goal specified.
In this thesis, we develop two motion variation techniques that automatically modifies motion captured data and generates goal-directed motion. We define behavior profile that shows the pattern of human action and this behavior profile is modified such that the resulting motion achieve the specified goal. And experiment was done to show the effectiveness of these technique using hand as a control point.