零空间边界Fisher分析法及其在人脸识别中的应用  

Null Space Marginal Fisher Analysis and Its Application in Face Recognition

作　　者：杨军[1] 刘妍丽[2] 

机构地区：[1]四川师范大学计算机科学学院,四川成都610101 [2]四川师范大学数学与软件学院,四川成都610101

出　　处：《西华大学学报（自然科学版）》2014年第1期60-64,共5页Journal of Xihua University:Natural Science Edition

基　　金：国家自然科学基金(60736046);973计划课题(2009CB320803);四川省教育厅资助科研项目(11ZB069);四川省重点实验室项目(PJ2012001)

摘　　要：边界Fisher分析(MFA)是一种有效的特征抽取方法,但在人脸识别的应用中会遭遇小样本问题。基于此,提出一种利用零空间法求解MFA优化准则的算法。该算法通过在MFA的类内散度矩阵的零空间中最大化MFA类间离散度得到最优投影向量,从而避免MFA方法所遇到的小样本问题,同时也保留了包含在类内散度矩阵零空间中的鉴别信息。在标准人脸库上的识别实验结果表明,该算法的识别率高于LDA和MFA,并且较容易选择其最优低维特征空间的维数。Marginal Fisher analysis （MFA） is an efficient linear projection technique for feature extraction. The major drawback of applying MFA to face recognition is that it often encounters the small sample size （SSS） problem. In this paper, a strategy based on null space for solving optimization criteria of MFA is proposed to avoid this issue. It maximizes the class scatter of training samples on null space of within-class scatter matrix （ Sw ） in MFA and reserves the discriminant information contained in null space of S~. The per- formance of this method is tested in both ORL and Yale face databases. Experimental resuhs show that this method is effective and a- chieves higher recognition rate than LDA and MFA. Moreover, it is easy to decide most optimal dimensionality of feature space for this method.

关 键 词：人脸识别 边界Fisher分析 小样本问题 零空间 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]