首页 | 本学科首页   官方微博 | 高级检索  
     

改进的完全二维主成分分析及其在步态识别中的应用研究*
引用本文:贲晛烨,安实,王科俊,王健. 改进的完全二维主成分分析及其在步态识别中的应用研究*[J]. 计算机应用研究, 2011, 28(6): 2088-2091. DOI: 10.3969/j.issn.1001-3695.2011.06.024
作者姓名:贲晛烨  安实  王科俊  王健
作者单位:1. 哈尔滨工业大学,交通科学与工程学院,哈尔滨,150090
2. 哈尔滨工程大学,自动化学院,哈尔滨,150001
基金项目:国家“863”计划资助项目(2008AA01Z148)
摘    要:对完全二维主成分分析算法进行改进,提出三种不同的加权策略,详细地分析它们的本质,并将其应用到步态识别中。在中科院自动化所提供的CASIA(B)步态数据库下验证加权方案的有效性,实验结果表明加权幂指数的选取对识别结果的影响比较大,通过实验可以选取最佳的权值,能够做到提高识别性能。最后针对各个行走状态下的步态,分析了背包步态识别率低的原因。

关 键 词:步态识别   步态能量图   完全二维主成分分析   加权完全二维主成分分析
收稿时间:2010-11-11
修稿时间:2011-05-15

Study on improved complete two dimensional principal component analysis and its application to gait recognition
BEN Xian-ye,AN Shi,WANG Ke-jun,WANG Jian. Study on improved complete two dimensional principal component analysis and its application to gait recognition[J]. Application Research of Computers, 2011, 28(6): 2088-2091. DOI: 10.3969/j.issn.1001-3695.2011.06.024
Authors:BEN Xian-ye  AN Shi  WANG Ke-jun  WANG Jian
Affiliation:(1.School of Transportation Science & Engineering, Harbin Institute of Technology, Harbin 150090, China; 2.College of Automation, Harbin Engineering University, Harbin 150001, China)
Abstract:This paper improved complete two dimensional principal component analysis, and proposed three different weighted strategies. Analyzed the essence of weighted scheme in detail. To evaluate the validity of the proposed method and apply it to gait recognition, conducted experiments on CASIA(B) gait database. Experimental results show that the selection of weighted power exponent has great influence on the recognition result, and can improve the recognition performance by picking the best weight value in experiments. In the end, discussed the reason why the recognition rate of gait with a bag was low in the various walking states.
Keywords:gait recognition   gait energy image   complete two dimensional principal component analysis(C2DPCA)   weighted complete two dimensional principal component analysis(WC2DPCA)
本文献已被 CNKI 万方数据 等数据库收录!
点击此处可从《计算机应用研究》浏览原始摘要信息
点击此处可从《计算机应用研究》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号