基于改进的二维线性判别分析的人脸识别

Face Recognition Based on Improved Two-dimensional Linear Discriminant Analysis

本文提出一种新的特征提取方法,人脸图像在2DPCA投影的基础上进行B2DLDA投影提取出人脸特征。这种方法克服了传统PCA和LDA方法的小样本问题和维数灾难问题,并且充分利用了二维人脸图像矩阵空间结构信息,大幅度降低了人脸特征维数。实验证明这种方法的识别率比传统的PCA和2DPCA方法高,识别时间和训练时间比传统的PCA和2DPCA方法少。

This paper proposed a new feature extraction method, that face feature is extracted based on projections on the 2DPCA and B2DLDA. This method overcomes the Small Sample Size problem and the curse of dimension of traditional PCA and LDA while using the two-dimensional information adequatedly and reducing the face feature dimension. Experiment shows that this method improves the classification accuracy and classification time and training time compared with the traditional PCA and 2DPCA method.