基于支持向量机方法的人脸识别研究

采用 SVM方法进行人脸识别研究 ,将人脸识别这一典型的多分类问题构造成适合 SVM处理的二分类问题 ,克服了传统 SVM方法在解决多分类问题上的一些缺陷 .实验以手工与自动两种预处理方式在 FERET和 Bio ID人脸库上完成 ,并与 PCA方法进行了对比 ,结果表明本文的 SVM方法比 PCA方法有更好的概括能力和更高的正确识别率 ,使得今后建立一个基于 SVM方法的人脸自动检测和识别系统成为可能

Face Recognition Based on Support Vector Machines

This paper uses SVM method to study face recognition, which is a typical problem of multi-classification. Only one SVM is built to solve this problem, thus overcomes some flaws of several traditional multi-classification methods of SVM. The experiments are done on the FERET and BioID face databases with manual as well as automatic means. The results show that this SVM method has better capability of generalization and higher rate of correct recognition than PCA method. As the method of face detection used in these experiments is also effective, the realization of an automatic face detection and recognition system based on SVM is possible in the future.