项目名称: 基于图模型的人体运动捕获数据检索研究
项目编号: No.61271362
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 无线电电子学、电信技术
项目作者: 肖秦琨
作者单位: 西安工业大学
项目金额: 65万元
中文摘要: 随着计算机动画技术的广泛使用以及运动捕获装置的普及,人体运动捕获数据急剧膨胀。对运动捕获数据进行高效编辑及重用的迫切需求,已使得多媒体检索领域面临新的挑战。运动捕获数据检索涉及复杂时间序列信号处理及数据存储结构设计,本项目在深入分析其特点的基础上,选用复杂系统建模的有力工具- - 图模型对其进行表达与处理,以使得检索效率最优化。本项目重点研究基于图模型的人体运动捕获数据检索问题,具体的研究内容包括:运动捕获数据基于内容和数值的表达方法研究,运动捕获数据基于内容和数值的相似性度量机制研究,运动捕获数据降维及数据存储方式研究等。在以上理论和关键技术研究的基础上,构建运动捕获数据基于多种形式查询的检索系统框架。本项目具有前瞻性和挑战性,其研究成果可用于电影特技制作、军事模拟训练、高水平运动员训练等领域。本研究在理论和关键技术上的突破对于探索复杂时间序列相关问题具有重要的理论意义和实用价值。
中文关键词: 运动数据;图模型;检索系统;运动匹配;动态贝叶斯网络
英文摘要: With computer animation and motion capture devices widely used, a large amount of human motion capture data is obtained. It is the urgent demand to edit and reuse motion capture data efficiently that makes multimedia retrieval technology face a new challenge.Based on in-depth analysis of human motion capture data, the research work will be particularly focused on human motion capture data retrieval, involving complex time series signal processing and data storage structure design. To optimize the retrieval efficiency, the graph model, which is a powerful tool for complex system modeling, is selected for expressing and processing human motion capture data. The key works of our proposed human motion capture data retrieval are: the content-based and numerical-based human motion description, the similarity measurement of content-based and numerical-based human motion, and the reduction of data dimension and data storage capacity. The whole retrieval system will be constructed based on the above-mentioned technologies to meet large variety of retrieval enquiries. Our proposed project can be prospectively applied to various applications,e.g., movie stunt production, military simulation training, high-level athletes training, etc. Furthermore, the exploration of complex time series processing and the relavant research
英文关键词: motion data;graph model;retrieval system;motion matching;dynamic Bayesian network