对称核主成分分析及其在人脸识别中的应用

Symmetric Kernel Principal Component Analysis and Its Application in Face Recognition

核主成分分析(KPCA)没有充分利用人脸的对称性特征,在人脸识别中缺少训练样本,致使其识别率较低。为此,提出一种对称KPCA算法。利用人脸的镜像对称性,通过对训练样本进行镜像变换,得到奇对称样本和偶对称样本,分别提取各奇/偶对称样本的特征分量,使用最近邻距离分类器完成分类。实验结果表明,该算法能扩大样本容量,当多项式阶数为2时,该算法的识别率高于KPCA算法,识别时间短于KPCA算法。

Aiming at the problem that Kernel Principal Component Analysis（KPCA） can not effectively use the feature of face symmetry, and generally lacks of training samples in face recognition, so the recognition rate is low. Therefore, this paper proposes a Symmetrical Kernel Principal Component Analysis（SKPCA） algorithm. This algorithm fully utilizes the face mirror symmetry, the odd symmetry samples and the even symmetry samples are received by mirror transforming for training samples. Odd/even symmetrical principal components are respectively extracted. A nearest neighbor classifier is employed to classify the extracted features. Experimental results show that this algorithm enlarges the number of training samples, when the polynomial order number is 2, the recognition rate of this algorithm is better than that of the KPCA algorithm, and recognition time is shorter than KPCA algorithm.