中文摘要：

针对低维控制信号不能直接生成精确的高维人体全身运动动画的难题,提出了一种基于窗口的线性回归模型,此模型利用提前捕获的运动数据库中自然的人体运动,在最大后验框架下和控制信号一起完成人体全身运动的精确重构.在用户给定的低维约束信号的控制下,通过对不同运动的生成验证了提出方法的有效性.通过给定一个合适的运动数据库,由本方法生成的结果能够和商业的运动捕获系统得到的结果相匹敌.

英文摘要：

Focusing on the issue that low-dimensional control signals can not directly construct precise high-dimensional entire human body motion,this paper introduces a window-based linear regression model,which works with control signals together,to accurately construct human motions under a maximum posterior framework.The proposed model is automatically created based on motion capture database at run time.We testified the effectiveness of proposed model by synthesizing different motions.If the database is good enough,the results of our system can be comparable with the results of expensive commercial mocap systems.