SPCA参数对单样本人脸识别效果影响分析  

Analysis of the influence of SPCA parameters on the recognition of a single sample face

作　　者：王科俊[1] 邹国锋 张洁[1] 

机构地区：[1]哈尔滨工程大学自动化学院,黑龙江哈尔滨150001

出　　处：《智能系统学报》2011年第6期531-538,共8页CAAI Transactions on Intelligent Systems

基　　金：国家"863"计划资助项目(2008AA01Z148);黑龙江省杰出青年科学基金资助项目(JC200703);哈尔滨市科技创新人才研究专项基金资助项目(2007RFXXG009)

摘　　要：奇异值扰动的主分量分析(SPCA)是一种有效的单样本人脸识别方法,但SPCA算法的识别效果受参数选择的影响比较大,针对SPCA算法中衍生图像生成参数n和结合参数α的不同取值对识别效果的影响进行了分析,利用ORL人脸库和CAS-PEAL人脸库做了大量的实验和比较分析,实验结果表明给出的SPCA参数选取方法和取值范围是合理的,并有效地提高了SPCA算法的实际应用效果和单样本人脸识别的性能.Singular value decomposition perturbation principal component analysis（SPCA） is an effective single-sample face recognition method;however,the identification results of the SPCA algorithm are seriously affected by parameter selection.In this paper,the effect on the identification,which was caused by the derived image parameter and the combined image generation parameter in the SPCA algorithm,was analyzed.Many experiments and comparative analyses were performed on the basis of the ORL face database and the CAS-PEAL face database.The experimental results show that the SPCA parameter selection method and the parameter range given in this paper are reasonable.In addition,reasonable parameters are effective in improving practical application of SPCA algorithms and the recognition performance of a single-sample face.

关 键 词：人脸识别 奇异值分解 结合投影主分量分析 奇异值扰动主分量分析 衍生图像 结合图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]