一种基于个人身份认证的正面人脸识别算法  

An Effective Approach for Frontal Face Verification

作　　者：陶亮[1] 庄镇泉[1] 

机构地区：[1]中国科学技术大学电子科学与技术系

出　　处：《中国图象图形学报（A辑）》2003年第8期860-865,共6页Journal of Image and Graphics

基　　金：安徽省自然科学基金项目 (0 10 42 2 10 );教育部优秀青年教师资助计划项目 (教人司 [2 0 0 2 ] 40号 );微软亚洲研究院与中国科学技术大学合作研究项目

摘　　要：利用小波分解提取人脸特征技术和支持向量机 (SVM)分类模型 ,提出了一种基于个人身份认证的正面人脸识别算法 (或称为人脸认证方法 ) .针对 M个用户的人脸认证算法包括二个阶段 :(1)训练阶段 :使用小波分解方法对脸像训练集中的人脸图象进行特征提取 ,并用所提取的人脸特征向量训练 M个 SVM(对应 M个用户 ) ;(2 )认证阶段 :先由待认证者所声称的用户身份 (姓名或密码等 )确定对应的一训练好的 SVM,然后用这一 SVM对小波分解方法提取的待认证人的脸像特征向量进行分类 ,分类结果将显示待认证人所声称的身份是否真实 .利用 ORL人脸图象库对该算法的实验测试结果 。This paper presents an effective algorithm for frontal face verification based on the wavelet decomposition technique and Support Vector Machines (SVMs). The process of the proposed method for face verification of M clients consists of two stages.(1) Training stage: by the wavelet decomposition, extracting the appropriate features from the facial images in the prepared training database of faces, training M SVMs for the M clients by the extracted facial feature vectors. (2) Verification stage: selecting a trained SVM from the M SVMs based on the identity claim (such as a name or a password) of an unknown person and using the trained SVM to classify the facial feature vector extracted from the facial image of the unknown person by the wavelet decomposition, and the classification result will show whether or not the identity claim of the unknown person is valid. The ORL database of faces is selected to test and evaluate the proposed algorithm. The results of the test are encouraging and the SVMs in the proposed algorithm are shown to perform very well in classification capability when compared to the traditional radial-basis function networks applied as classifiers in the algorithm.

关 键 词：个人身份认证 小波分解 正面人脸识别算法 神经网络 支持向量机 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]