Editing and generation of realistic motions of human body is considered as one of the challenging problems in Animation and Computer Graphics. One of the fundamental approach is capturing motion from real actors. Motion Graphs is one of the key methods to combine these captured motions into a longer motion. A drawback of Motion Graphs is that motions with different momentums cannot be blended or can be blended poorly. In order to address this problem, we aim to use a dynamics based parameterization approach. Full dynamic solutions and motion correction methods have already been proposed in the literature. Although these approaches are physically valid, their calculation costs are fairly high. In our approach, pre-computed set of parameters will be used to generate visually realistic transitions to improve online calculation costs. First, a set of basic motion patterns will be selected. For each motion pattern some pre-suggested set of parameters will be used. The base of these parameters will be selected manually. By use of motion captured data, the behavior of these parameters will be observed. Then, we will try to apply these learned behaviors to probable motion transitions, which are not existent due to high momentum differences. A weighted combination of these learned parameters for each motion pattern will be used to generate transitions for the motions that are out of this selected set. Other than providing a method for combining motions with different momentums, we will test our solution in path finding algorithms, in which use of motion graphs with insufficient transitions is considered as a problematic issue.

DATE: 7 April, 2008, Monday@ 15:30
PLACE: EA 409