项目名称: 草绘交互接口下基于精简存储的人体运动数据检索技术研究
项目编号: No.61300089
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 自动化技术、计算机技术
项目作者: 王鹏杰
作者单位: 大连民族学院
项目金额: 25万元
中文摘要: 数据驱动的角色动画技术和触屏技术的快速发展使得基于草绘接口的实时运动检索成为一个亟待研究的问题;同时,传统运动数据压缩和检索方法相互隔离,直接整合的算法框架存在计算复杂、数据冗余度高的缺点。针对上述问题,本项目研究草绘接口下基于多组几何特征的运动检索方法,定义静态和动态几何特征,对运动数据进行局部匹配和粗糙/精细粒度控制。接着,研究运动数据及特征向量精简压缩组织的方法,基于逆运动学对预测参数进行反馈调整,对数据进行无损压缩和实时解压。根据上述两点,从运动数据库组织全局出发,提出了运动压缩/检索方法有机结合思路:基于"动作纹理"分析,研究统一运动压缩和检索的运动数据库组织模型;在此基础上,得到基于精简结构的运动检索方法,使得根据用户输入,能实时从无损压缩的运动数据库中检索出符合条件的运动片段。本项目为运动数据精简存储及基于其上的草绘检索提供一条新途径,有望促进动画游戏、虚拟现实等产业的发展。
中文关键词: 角色动画;运动捕获;运动压缩;运动检索;草绘接口
英文摘要: The development of data-driven character animation and touch screen technology has driven sketch-based real-time retrieval of motion data to become a new research topic. On the other hand, motion compression and motion retrieval are considered separately. The naive method of putting the two together will be time-consuming and bring lots of redundancy. In this proposal, we first propose a sketch-based motion retrieval method based on the multi-group geometry features. By defining pose and motion geometry features, the method is capable of partial matching, with course/detail level control. We then propose a lossless compression method to compress the motion data and feature vectors. This method is based on the adjusting the prediction parameters using inverse kinematics feedbacks and can decompress in real-time. Finally, we propose the concept of unifying motion compression/retrieval. Based on "action texture" analysis, we explore how to organize motion capture database and propose a motion retrieval method based on compact storage structure. Users can retrieve motion segments in real-time from a losslessly compressed motion database. This proposal provides a new solution to address the problems of how to compactly store motion data and how to retrieve the motion data from this storage structure based on sketchin
英文关键词: Character animation;Motion capture;Motion compression;Motion retrieval;Sketch interface