一种局部敏感的核稀疏表示分类算法  

A classification algorithm based on locality-sensitive kernel sparse representation for face recognition

作　　者：张石清[1] 赵小明[1] 楼宋江[1] 闯跃龙[1] 郭文平[1] 陈盈[1] 

机构地区：[1]台州学院图像处理与模式识别研究所,浙江台州318000

出　　处：《光电子．激光》2014年第9期1812-1817,共6页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金(61272261;61203257)资助项目

摘　　要：为了克服核稀疏表示分类(KSRC)算法无法获取数据的局部性信息从而导致获取的稀疏表示系数判别性受到限制的不足,提出一种局部敏感的KSRC(LS-KSRC)算法用于人脸识别。通过在核特征空间中同时集成稀疏性和数据局部性信息,从而获取具有良好判别性的用于分类的稀疏表示系数。在标准的ORL人脸数据库和Extended Yale B人脸数据库的试验结果表明,本文方法的分类性能优于传统的(KSRC)算法、稀疏表示分类(SRC)算法、局部线性约束编码(LLC)、支持向量机(SVM)、最近邻法(NN)以及最近邻子空间法(NS),用于人脸识别能够取得优越的分类性能。Kernel sparse representation-based classification （KSRC） is currently one of hot research topics in the fields of pattern recognition and computer vision KSRC has excellent classification performance, but is not able to obtain the important information about data locality,resulting in the fact that the discriminating power of sparse representation coefficients yielded by KSRC is reduced. Considering the importance of data locality, in this paper a new classification algorithm based on locality-sensitive kernel sparse representation is proposed for face recognition. The proposed method integrates both sparsity and data locality in the kernel feature space so that it can obtain good discriminating sparse representation coefficients for classification. Experimental results on two benchrnarking face databases, i. e. , the ORL database and the Extended Yale B database, demonstrate that the proposed method outperforms kernel sparse representation-based classification （KSRC）, sparse representation-based classification （SRC）,locality-constrained linear coding （LLC）, support vector machines （SVM）, the nearest neighbor （NN）, and the nearest subspace （NS）. Therefore,the proposed method is able to achieve promising performance for face recognition.

关 键 词：核稀疏表示 数据局部性 人脸识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]