核方法在人脸识别中的应用

1 引言人脸识别技术广泛应用于身份验证、门检系统以及人员监视等方面,在过去的几年里,人脸识别技术有了很大的发展。人脸识别技术与普通的模式识别不同,主要是因为在一般的模式识别中,有几个分类,每个分类中有很多样本,这样可以安排大量样本进行训练;相反,人脸识别中通常会有很多不同的人脸,每个人脸代表一个分类,而每个分类中的样本数都比较少,在很多情况下,甚至每个人只有一张图片(如身份证照片),在文[4]中提出了处理只有一个样本情况下的人脸识别。

Face Recognition Using Kernel Methods

Kernel function is the function which computes dot product in feature spaces. Both the SVMs and kernel PCA are kernel.based learning methods. In this paper, the SVMs and kernel PCA are used to tackle the face recognition problem. SVMs are classifiers which have demonstrated high generalization capabilities. Kernel PCA is a feature extraction technique which is proposed as a nonlinear extension of a PCA. We illustrate the potential of SVMs and kernel PCA on the Yale database and compare with a PCA based algorithm. The experiments indicate that SVMs and kernel PCA are superior to the PCA method.