Generating Character Motion And Posture Based On Low-dimensional Feature

The wide application of motion capture equipment brings more and more motion data, although these data can provide sufficient support for human motion generation, the large amount of data also brings a series of difficult problems in classification, marking, retrieval, storage and so on. In addition, the topological structure of the human body is so complex and too many controllable variables are involved, that it becomes a very difficult problem for directly controlling human movement through manipulating these high dimensional variables. To solve these problems, we focus on the organization, dimensionality reduction and search for motion data. Based on corresponding low-dimensional feature space, we study on the approaches for character motion and posture concerning to reaching and sketching-based system.To overcome the defects of flexibility and realism for existing algorithms for reaching motion synthesis, we present a Biomechanics-based Life-like Reaching Controller (BLRC). BLRC employs various reaching strategies borrowed from biomechanics and exploits the arm-reachable workspace to guide the motion sampling. Through constructing different low-dimensional spaces for reaching strategies, the BLRC can solve the optimization problem in the low-dimensional space quickly.To further improve the realism of reaching motion, we present a controller-based framework for reaching motion synthesis. Based on low-dimensional motion feature, we construct four stationary controllers to generate concrete reaching motion and three transition controllers to stitch these stationary controllers. The experiments show that our framework can express the inherent tentative process for human reaching effectively.In order to facilitate users to create a right human posture quickly, we create a user-friendly sketch-based character posing system. By mapping the high-dimensional image descritptor into low-dimensional space, our system realizes the fast retrieval of3D character pose in a large database. While the user is sketching, our system is able to provide some feedback information to help him/her to draw next stroke.