Abstract: A face recognition method based on 2D Principal Component Analysis(2DPCA) and compressive sensing is introduced in this paper. The 2DPCA is used to obtain the feature vectors and the compressive sensing is used to get the sparse representation of the test image given enough training images. The measure matrix is composed of the feature vectors of all training images and the feature vector of each test image is the observation value, then the encoding signal of the test image, which will be used for face recognition, can be obtained by using the L1 norm minimization. Experimental results indicate that the recognition rate based on 2DPCA and compressive sensing is higher than the recognition rate using other methods.

Key words: face recognition, compressive sensing, 2D Principal Component Analysis(2DPCA), L1 norm, sparse representation

摘要： 提出一种基于二维主成份分析(2DPCA)和压缩感知的人脸识别方法。阐述2DPCA提取特征向量的工作原理，利用压缩感知方法求解待识别图像在足够样本下的稀疏表示。由所有训练图的特征向量构成测量矩阵，将每一幅待识别图像的特征向量作为测量值，由压缩感知中求解的L1范数极小值得到待识别图像的编码信号，根据该编码信号识别人脸图像。实验结果表明，与其他组合方法相比，基于 2DPCA和压缩感知的人脸识别方法得到的识别率较高。

关键词: 人脸识别, 压缩感知, 二维主成份分析, L1范数, 稀疏表示

CLC Number: 

• TP242.6