摘要
提出了一种局部非参数子空间分析算法(Local Nonparametric Subspace Analysis,LNSA),将其应用在人脸识别中。LNSA算法结合了非参数子空间算法(Nonparametric Subspace Analysis,NSA)与局部保留投影算法(Locality Preserving Projection,LPP)。它利用LPP算法中的相似度矩阵重构NSA的类内散度矩阵,使得在最大化类间散度矩阵的同时保留了类的局部结构。在ORL人脸库和XM2VTS人脸库上作了实验并证明LNSA方法要优于其他方法。
A local nonparametric subspace analysis algorithm is proposed and applied to face recognition. The algorithm, which combines nonparametric subspace analysis with locality preserving projection and reconstructs the within-class scatter matrix by the affinity matrix of locality preserving projection algorithm, makes it possible to maximize the between-class scatter matrix and meanwhile to preserve the class local structure.The experimental results on ORL and XM2VTS face data-bases show that the performance of local nonparametric discriminant analysis is better than other algorithms.
出处
《计算机工程与应用》
CSCD
2014年第3期141-144,共4页
Computer Engineering and Applications
基金
江苏省科研创新计划项目(No.CXLX11_04910)
中央高校基本科研业务费专项资金资助(No.JUSRT211A70)
关键词
人脸识别
非参数子空间分析
局部保留投影
局部鉴别分析
局部非参数子空间分析
face recognition
nonparametric subspace analysis
locality preserving projection
local Fisher discriminant analysis
local nonparametric subspace analysis
