基于支持向量机的人脸识别方法  

Face Recognition Using Support Vector Machines

作　　者：崔国勤[1] 李锦涛[1] 高文[1] 焦锋[1] 

机构地区：[1]中国科学院计算技术研究所数字化技术研究室,北京100080

出　　处：《计算机科学》2003年第4期11-15,共5页Computer Science

基　　金：国家863计划"生物特征识别核心技术与关键问题研究"(项目编号:2001AA114190)

摘　　要：Support Vector Machines are a binary classification method and have demonstrated excellent results in pattern recognition. Face recognition is a multi-class problem, where the number of classes is of the known individuals. In this paper we use face data extracted from Eigenfeatures and develope a method to extend SVM to using in multi-class. The training set consists of 5 images of each of the 50 persons equally distributed among frontal, approximately 15°rotated respectively, and the test set consists of 10 images each of the 50 persons. In the ICT-YC face gallery, the proposed system obtains competitive results highly: a correct recognition rate of 94.8％ for all the 50 persons, to the less number of the persons and to the famous ORL face gallery we also get good face recognition rate.Support Vector Machines are a binary classification method and have demonstrated excellent results in pattern recognition. Face recognition is a multi-class problem, where the number of classes is of the known individuals. In this paper we use face data extracted from Eigenfeatures and develope a method to extend SVM to using in multi-class. The training set consists of 5 images of each of the 50 persons equally distributed among frontal, approximately 15°rotated respectively, and the test set consists of 10 images each of the 50 persons. In the ICT-YC face gallery, the proposed system obtains competitive results highly: a correct recognition rate of 94.8% for all the 50 persons, to the less number of the persons and to the famous ORL face gallery we also get good face recognition rate.

关 键 词：人脸识别 支持向量机 自动识别系统 人脸图像 计算机 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]