Abstract: This paper proposes a method of face recognition combining GLRAM（Generalized Low Rank Approximations of Matrices） with LDA（Linear Discriminant Analysis）.Firstly，effective feature can be obtained using GLRAM，and then LDA is used to depress the feature dimension and acquire the best classification feature.This enhances the discriminatory power of extracted features.Experimental results demonstrate that higher recognition rate can be achieved in shorter time.This proposed method outperforms GLRAM and LDA methods.

Key words: face recognition, Generalized Low Rank Approximations of Matrices（GLRAM）, Linear Discriminant Analysis（LDA）

摘要： 提出了一种GLRAM（矩阵的广义低秩逼近）与LDA（线性判别分析）相结合的人脸识别方法。首先利用GLRAM方法获得人脸图像的有效特征，然后通过LDA对获得的特征进行进一步的降维并获得最佳分类特征。这样使得抽取特征的判断能力得到了显著增强。实验结果表明，该算法在较短的时间内取得了较高的识别率，效果优于单独运用GLRAM方法和LDA方法。

关键词: 人脸识别, 矩阵的广义低秩逼近（GLRAM）, 线性判别分析（LDA）