基于局部一致性和相干性字典学习的人脸识别  

Local consistency and coherence based dictionary learning algorithm for face recognition

作　　者：张凤莉[1] ZHANG Feng-li(Department of Electrical and Mechanical Engineering,Shangqiu Polytechnic,Shangqiu 476000,China)

机构地区：[1]商丘职业技术学院机电工程系,河南商丘476000

出　　处：《电子设计工程》2019年第9期127-130,135,共5页Electronic Design Engineering

摘　　要：字典学习算法被广泛的用于人脸识别。字典的局部一致性及相干性对于字典学习很重要,然而当前很多字典学习算法都没有考虑到字典间的局部一致性,更没有同时将二者用于字典学习,致使学习到的字典的鉴别力不是特别强,进而使得它们的识别性能不是非常理想。针对这个问题,本文提出了一种基于局部一致性和相干性的字典学习算法,并将它用于人脸识别。本算法通过构造局部一致项和相干项学习具有很强鉴别力的字典,并利用学习到的字典重构样本,最终利用重构样本与测试样本间的残差完成分类任务,在AR,ORL和FEI人脸数据库上的实验结果表明本文提出的算法取得了不错的识别效果。Dictionary learning algorithms are widely used for face recognition. The local consistency and coherence of the dictionaries are important for dictionary learning. However,many dictionary learning algorithms do not consider the local consistency between dictionaries,and do not use them together for dictionary learning. As a result,the learned dictionaries are not very discriminative,which makes their recognition performance not very ideal.To solve this problem,this paper proposes a dictionary learning algorithm based on local consistency and coherence,and uses it for face recognition. This algorithm learns adictionary with strong discriminative ability by constructing local consistent term andcoherent term,and reconstructs the samples using the learned dictionary.Finally,the classification task is completed by using the residual between the reconstructed sample and the test sample.Experimental results on AR,ORL and FEI face databases show that the proposed algorithm achieves a good recognition effect.

关 键 词：模式识别 人脸识别 字典学习 局部一致性 相干性 

分 类 号：TN919.8[电子电信—通信与信息系统]