推广的PCA及其在人脸识别中的应用.pdfVIP

叠i酉圈垦簟鬯翦簟薯鬯_簟幢囊篓鋈鲎篓隧!!!竺璺望!!!!!!曼!塑旦璺21塑!!!!竺!型堕!!!!!!堕旦Y 推广的PCA及其在人脸识别中的应用 黄防，刘金琨，张宝昌 (北京航空航天大学自动化与电气工程学院，北京100191) 摘 像矩阵集进行分块，再利用传统PCA对分块得到的子训练矩阵集进行分析，得到多个变换矩阵，通过 这些变换矩阵将训练图片和测试图片投影到特征空间进行鉴别。与传统PCA方法相比，提高了主元 明显优于传统的PCA方法，识别率得到了提高。 关键词：主成分分析；特征抽取；推广的PCA；特征矩阵；人脸识别 中图分类号：TP391 文献标识码：A PromotedPCAandits inhumanface application recognition HUANG Bao JinKun，ZHANG Fang，LIU Chang ofAutomationScienceandElectrical (School Engineering，BeihangUniversity，Beijing100191，China) Abstract：Promotedhumanface based()n inthis PCA，a recognitiontechnique PCA，ispresented original paper．Firsdy，the is somesub—sets．Secondthe set dividedinto well-knownPCAmethodis lothesub-setsobtainedcorre— training directl)，used transformed these setand setare in feature sponding matrixes．Throughmatrixes，trainingtesting mapped space，doing withtraditional of at of dimension．Totest PCA，this improves tion．compared approach accuracyrecognitionthe[)ricehigher pro— of motcdPCAandtoevaluateits series are onahumanface databases：Ferethuman performance，aexperimentsperformed image indicatethe facedatabases．The results that of PCAis tothatof experimental performancepromoted obviouslysuperior PCA．The rateis recognitionimproved．