基于两空间核鉴别分析的人脸识别

指出了几种线性鉴别分析方法在处理小样本人脸识别问题时存在的不足,结合核方法的思想,提出了一种基于两空间核鉴别分析的人脸识别方法.首先使用KPCA方法在核变换后的特征空间中对样本进行处理;进面将变换后的类内散布矩阵分成非零空间和零空间进行鉴别向量确定和鉴别特征提取,最后将得到的两种鉴别特征融合,从而使用最近邻法进行分类....

Face recognition based on two subspaces kernel discriminant analysis

Disadvantages of several typical variants on LDA while dealing with the well- known small sample size problem in face recognition are revealed. Based on the kernel idea＇ a new discriminant analysis method named two subspaces kernel discriminant analysis is proposed to deal with small sample size problem in face recognition. Firstly, all the samples are processed in the feature space generated by KPCA. Secondly, discriminant analysis is carried out in the zero subspace and nonzero subspace of the within- class scatter matrix respectively, and two kinds of discriminant features are obtained. Thirdly, the two kinds of discriminant features are fused to determine class of the sanles. The method can not only process small sample size face recognition but also extract discriminant features in complex conditions effectively.Experimental results show that the proposed algorithm is superior to linear discriminant analysis methods and traditional kernel methods in simplifying face distribution and classifying faces.