Abstract: As the nonlinear extensions of Principal Component Analysis（PCA）, Kernel Principal Component Analysis（KPCA） is effective for face recognition. In order to reduce recognition time, a face recognition method based on Discrete Cosine Transform（DCT） and KPCA is presented. The feature coefficients are extracted by DCT, and part of the coefficients are chosen to reconstruct face images. The face feature of high dimention is extracted by KPCA. The nearest neighbor classifier is used for identification. The experiment result on ORL face databases shows this method has the property of being faster, and the comprehensive performance is better than that of KPCA.

Key words: face recognition, feature extract, Kernel Principle Component Analysis（KPCA）, Discrete Cosine Transform（DCT）, nearest neighbor classifier

摘要： 核主成分分析是主成分分析在核空间中的非线性推广，能有效应用于人脸识别，但是识别过程时间开销过大仍是待解决的问题。提出了一种结合离散余弦变换和核主分量分析的人脸识别方法。对人脸图像进行离散余弦变换，选择部分系数重建图像，采用核主分量分析的方法提取人脸特征，采用最近邻分类器进行识别。在ORL人脸库上的仿真结果表明所提出的方法速度快，综合性能优于核主成分分析方法。

关键词: 人脸识别, 特征提取, 核主成分分析, 离散余弦变换, 最近邻分类器