Abstract
Recent advances in motion capture techniques facilitate the active research in example-based human motion synthesis. However, many proposed solutions only identify low-level motion elements which are derived from mathematic analysis of motion examples. These solutions provide no direct connection between high-level motion specification and low-level motion elements. This drawback limits the usefulness of example-based approaches. To alleviate the above mentioned gap, this study proposes some possible solutions to the high-level motion synthesis and control. The proposed solutions emphasize on the realization processes of text-based motion synthesis and quality-based style synthesis. For text-based solution, we use text analysis method to identify a set of meaningful basic motions. Then, an annotation method is introduced to automatically label motion data according to given basic motion texts. Therefore, a new motion with given textual description can be synthesized by concatenating motion clips of basic motion texts in the description. As for style synthesis, we use motion qualities in Laban Movement Analysis (LMA) to construct our quality-based style synthesis kernel. As the Effort component of LMA has been used by many movement related fields, we present a dynamics-based Effort simulator so that a given motion capture data can be modified according to specified Effort qualities. The basic idea of our Effort simulator is to establish relations between Effort factors such as Space, Weight, Time, and Flow, and their corresponding dynamics parameters such as force, stiffness, inertia, damping, gravity, and etc. Then, a specified Effort quality can be realized by giving proper values of corresponding dynamics parameters. In other words, the proposed simulator not only can mimic the given motion effectively, but also provides control capability for varying given movement according to specified Effort qualities. Moreover, we also propose solutions to constraint satisfiable problem so that motion details and dynamics constraints of the original motion capture data can be preserved after modification.