广义的鉴别局部中值保持投影及人脸识别  

Generalized Discriminant Local Median Preserving Projections and Face Recognition

作　　者：张永[1] 万鸣华[1,2] 

机构地区：[1]南昌航空大学信息工程学院,南昌330063 [2]南京审计大学工学院,南京211815

出　　处：《计算机科学》2018年第2期90-93,108,共5页Computer Science

基　　金：国家自然科学基金项目(61462064); 中国博士后基金项目(2016M600674); 江苏省自然科学基金面上项目(BK20161580)资助

摘　　要：针对鉴别的局部中值保持投影(DLMPP)在小样本情况下面临的类内散布矩阵奇异的问题,提出了广义的鉴别局部中值保持投影(GDLMPP)算法。GDLMPP首先将样本等价映射到一个低维子空间,然后在此子空间求解最佳投影矩阵,从而有效解决了小样本问题,并从理论上验证了当类内散布矩阵非奇异时,GDLMPP等价于DLMPP。最后,通过在ORL及AR库上的实验验证了算法的有效性。To solve the problem of the singularity of the within-class scatter matrix in discriminant local median preserving projections（DLMPP）in the case of small sample problem,an algorithm named generalized local median preserving projection（GDLMPP）was proposed.To solve the small sample problem,GDLMPP firstly transforms the samples into a lower dimensional space equivalently,and then solves the optimal projection matrix.The theoretical analysis shows that GDLMPP is equivalent to DLMPP when the within-class scatter matrix is non-singular.At last,the experimental results validate the effectiveness of the proposed algorithm on the ORL and AR face databases.

关 键 词：人脸识别 特征提取 小样本问题 鉴别的局部中值保持投影 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]