基于多源信息融合的人体运动分析与建模研究-模式识别与智能系统专业论文.docxVIP

中国科举技术大学博士论文嚣褥究分群，在大量获淑逮西爱力蓓惑熬基疆上，结合运麓生物力学魏知 中国科举技术大学博士论文 嚣褥究分群，在大量获淑逮西爱力蓓惑熬基疆上，结合运麓生物力学魏知 识分析了遮些力学行为的地面反力特征，并进行了特征的提取，首次引入 支持蠢量橇(SVM)等方法，褥遗了，j、样本绩戮下静邃凌模式疆弱分类器， 取得了满意的分类效果。 {；i菰举运动为弼，对运动过程孛熬瓶电信号、逡蕊反力等多滚僚惑， 基于多输入和多输出(MIMO)支持向殛机和神经网络集成方法，谶行了 飘电信息灏薜力信息麓会的运动阶段分解与谈掰豹磊芽究，撬商了运渤阶段 分类器的貔棒性，降低了分类的错误率。 建立了探举遮动提铃阶段的入体摸鹜，邋遗模鹫酋次磷究了捷铃阶段 力(力矩)随时间变化的规律，为运动损伤的预防和改进技术提供参考。 最后，针对传统建模存程的问题，尝试了用多源信息融合来估计关节力矩 的新方法，得到了初步研究结果。 关键词： 信息融合 运动生物力学人体运动仿真 举重运动神经 网络支持向量帆 模式识别 中国科学技术大学博}论文Abstract 中国科学技术大学博}论文 Abstract With the development of sports scientific research，the computer simulation and modeling of human body for diagnosing athletes’technique has been an irreversible trend．It has great value in theory and practice to simulate the human body motion of athletes with the knowledge of sports biomechanics．Much interesting and elaborate research had been done by scholars in many fields at home or overseas，Newton motion law is cited as the foundation of the traditional solution of human body modeling．By simplifying the configuration of human body，listing the motion equations， setting the initializing condition etc，the simulating results can be approximately obtained．Computing error can’t be avoided more or less using the above method，where simplified modeling is adopted，and difficulties also exist in the modifying the modeling．The new viewpoint of human body motion analyzing and modeling based on the fusion of multi—information is proposed in the paper by systematically investigating the means of current human body modeling and simulating．Beginning with the movement information acquisition and analysis，many mathematics tools are utilized tO extract the features of movement information such as the information of ground reaction force，surface electromyographic(SEMG)signals and motion image resolution．Taking weight lifting as the example，the automatic decomposition and recognition of the movement pattern is implemented via multi—information fusion．The model of weight lifting is constructed with the