基于模糊模式识别的人体姿态识别  

Human posture identification based on fuzzy pattern recognition

作　　者：王玉坤[1] 高炜欣[1] 汤楠[1] 王欣[1] 姬江 

机构地区：[1]西安石油大学电子工程学院,陕西西安710065 [2]西安西北石油管道公司,陕西西安710065

出　　处：《计算机工程与设计》2016年第6期1621-1625,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(E040901);2013陕西省自然科学基金项目(2013JQ8049);陕西省自然科学基础研究计划青年人才基金项目(2015JQ5129);中国石油科技创新基金研究基金项目(2014D-5006-0605)

摘　　要：针对现有的人体姿态识别算法,采用一种模糊数学的方法对人体姿态进行识别。提取一些常用的时域特征作为姿态识别的特征量,通过重力学原理及对步态的分析,提出一种加速度数据特征Me-z,突出姿态之间数据的差异和分布特点。采用基于择近原则的模糊模式识别的方法对人体姿态进行识别,通过ROC曲线及测度矩阵的方法对分类结果进行精度评价,验证该方法的可行性。实验结果表明,该算法为人体姿态识别提供了一种有效的方法。An algorithm of human activity pattern recognition based on fuzzy mathematics was proposed for the existing ones.Some time-domain features and a feature Me-z were proposed through applying gait analysis and the principle of gravity rate to activity pattern recognition,and the difference and distribution of acceleration data were highlighted.The fuzzy pattern recognition method based on the principle of fuzzy closeness optimization was adopted to posture recognition,and the precision evaluation was carried out for the classified results through ROC curve and measurement matrix,which confirmed the feasibility of this algorithm.The results show that the algorithm is an effective method for human posture recognition.

关 键 词：加速度 时域特征 模糊 混淆矩阵 相似性 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]