摘要
基于动作的视频交互游戏一直是游戏市场上非常受消费者青睐的主流游戏之一。研究利用Kinect从用户动作中获取低维控制信号,然后通过双级结构来重建高维度运动控制信号,以实现高质量人体动画的实时合成。其中第一级先通过构造一个邻居图缩小搜索空间,再通过K-D树加速搜索算法得到k个相似数据,最后基于主成分分析法来构建一个线性运动实时合成模型;第二级则是利用平滑参数对线性模型进行优化。实验结果表明,即使在场景受到严重干扰的情况下,该方法仍然可以重建出高质量的人体动画。
The interactive video game based on human action is always one of the most popular games on the game market. This research gets low-dimensional control signals from uses＇ actions, then rebuilds high-dimensional motion control signals through two-level structure to realize quality real-time synthesis of human body animation. The first level is to get k similar data by constructing an adjacency graph for a smaller searching space using a K-D tree accelerating searching algorithm, until finally construct a real-time linear motion synthesis mode. The second level is to optimize the mode by smoothing parameters. Experimental results show that even when the scene is interfered seriously, the method can still rebuild quality human body animations.
出处
《计算机应用与软件》
CSCD
2011年第11期184-187,共4页
Computer Applications and Software
基金
浙江省自然科学基金(Y1110882)
浙江省教育厅科研项目(Y200907765
Y201018160)
