基于特征加权的人脸识别  

Face recognition based on feature weighted and performance comparison

作　　者：朱玉莲[1] 

机构地区：[1]南京航空航天大学计算中心,江苏南京211000

出　　处：《计算机应用》2005年第11期2584-2585,共2页journal of Computer Applications

摘　　要：现有的人脸识别方法通常未考虑不同特征或像素对识别结果的影响。实际上,人脸面部不同特征在人脸识别过程中的作用是不同的。研究了各个特征在识别中的作用,分别采用三种加权方法对人脸图像进行了预处理,并应用流行的人脸识别方法(联想记忆、主分量分析和Fisher线性判别分析)进行识别。最后用标准人脸库ORL进行了实验,实验结果表明特征加权方法对人脸识别是有效且通用的。Current face recognition methods usually do not consider the effects of different facial features. After the roles of different facial features in the course of face recognition were researched, face images were preprocessed by three feature weighted methods respectively, then were recognized by popular face recognition methods of association memory, principal component analysis and Fisher linear discriminant analysis. The experiments on ORL face database show that feature weighted methods are effective and general to face recognition.

关 键 词：人脸识别 联想记忆 主分量分析 FISHER线性判别分析 特征(混合)加权 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]