基于多分类器融合的人脸识别方法  

Face Recognition Based on the Fusion of Multiple Classifiers

作　　者：陈羽[1] 赖剑煌[1] 

机构地区：[1]中山大学电子通信工程系

出　　处：《中山大学学报（自然科学版）》2006年第4期24-27,共4页Acta Scientiarum Naturalium Universitatis Sunyatseni

基　　金：国家自然科学基金资助项目(60373082);教育部科学技术重点项目资助项目(105134)

摘　　要：提出了一种融合整体和局部信息进行人脸识别的新方法。首先利用DCT+LDA方法提取表达人脸信息能力强的左眼、右眼和嘴巴的局部特征,利用F isherface方法和简单频谱脸方法提取人脸的整体特征,然后应用多分类器组合规则融合整体和局部特征,实验结果表明利用加法融合规则在ORL和FERET数据库上识别率分别达到98.45%和90.79%,说明了该方法的有效性,同时也表明将多分类组合应用于人脸识别是一种比较可行的思路。A new face recognition method applying the information of the global and local features is presented. First the best representation of local features of left eye, right eye and mouth is extracted with face recognition algorithms DCT ＋ LDA method and the global feature with algorithms Fisherface and simple spretroface method, and then the global and local features are combined applying the multiple classifiers fusion rule. Experimental results show the recognition rate is 98.45% in the ORL database and 90. 79% in the FERET database applying weigthed sum rule, showing the method is availably and the thought is feasible using multiple classifiers fusion for face recognition.

关 键 词：人脸识别 局部和整体特征 多分类器融合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]