Techniques For Motion Capture Data Based Animation

With the popularity of Motion Capture Systems since 1990s, Computer Animation has made great progress, and has been widely used in film, game, education and military. Problems, however, occur concerning how to produce animation efficiently and effectively using motion capture data. Under this background, motion capture data based animation has been a hot topic in Computer Graphics recently.The work of this thesis is to explore new approaches to motion capture data based animation, and presents the following algorithms, including wavelet transformation based motion synthesis and realism processing, example based motion retrieval, Markov process based motion graph, and multi-agents based group animation. The thesis is organized as follows:In the first chapter, we introduce the motivation, the state of the art and the brief description of this work.In the second chapter, we make a brief survey on motion capture technologies and motion capture data based animation.In the third chapter, we present an approach to editing motion both in frequency domain and time-spatial domain. Firstly, we introduce wavelet transformation into motion multi-resolution analysis and propose some new algorithms, namely motion feature enhancement, motion fusion and motion feature texture. Secondly, we propose a space-time constraint based motion rectification to preserve the reality of motion.In the fourth chapter, we present an example based 3D motion retrieval algorithm. Motion index tree is constructed based on the hierarchical motion description using dynamic clustering. The example motion is firstly classified using kNN algorithm according to the motion index tree. Then the similarity between the example and motions in the sub-library is calculated through elastic matching. To improve the efficiency, clustering based key frame extraction algorithm is adopted to reduce the feature dimensions.In the fifth chapter, we present a framework for generating motions from motion capture data. We segment each original motion into a motion primitive sequence, with each motion primitive defining a fundamental dynamics of the motion. Motion primitives sharing similar dynamics are clustered into a motion cluster. We model each motion as a first-order Markov process, with its state being a motion cluster. We construct a directed graph, called motion graph, to encapsulate the connection between motion clusters. New motion can be generated by synthesizing motion primitives along a constrained or specified path in the motion graph. Within this framework, we propose two promising techniques: random motion sampling and motion path synthesis.In the sixth chapter, we present a multiple autonomous agents based group animation framework. Each animated character in a group is modeled as an autonomous agent with perception, behavior planning and motion generation functions. As compared with traditional intelligent animated character, we generate motions for the animated character using motion capture data, instead of modeling the complex motion generation mechanism. The animated characters manage their movement autonomically. To produce the required group animation, animators only need to specify some simple information and record the useful group motion.In the seventh chapter, we conclude the work and discuss future work.