摘要
在基于稀疏表示的人脸鉴别方法中,提高鉴别准确率的关键在于增强字典和稀疏编码的辨别性.针对小样本训练情况,本文提出一种新的混合字典学习方法.首先以费舍尔判别准则和拉普拉斯矩阵为约束,利用类别特色字典提取数据类别之间的特殊性,在保留稀疏编码数据相似性的同时减小类内编码离散度,增大类间编码离散度.然后利用类内差异字典提取类别共性,捕捉不同类别的相同特征.最后将类别特色字典与类内差异字典相结合,分为4个实验方案在AR、CMU-PIE、LFW等人脸数据库上进行实验,结果表明该算法在少样本训练条件下可以获得更高识别精度.
To improve the accuracy of face recognition based on sparse representation,the key lies in discrimination of both dictionary and sparse coding.Aiming at the training of small samples,In this paper,we proposes a new hybrid dictionary learning method.Specifically,we first develop a discriminative class-specific dictionary amplifying the differences between training classes.The Laplacian matrix and Fisher criterion are integrated into the same model to preserve the similarity of sparse coding data while reducing the intraclass coding dispersion and increasing the inter-class coding dispersion.Secondly,we construct an intra-class variation dictionary to extract the similarity of classes and capture the same features of different classes.Finally,combining the class-specific dictionary with the intra-class difference dictionary,we divide into four experimental schemes and conduct experiments on AR,cmu-pie,LFW databases.The results show that the algorithm can obtain higher recognition accuracy under the condition of fewer samples training.
作者
矫慧文
狄岚
梁久祯
JIAO Hui-wen;DI Lan;LIANG Jiu-zhen(School of Artificial Intelligence and Computer Science,Jiangnan University,Wuxi 214122,China;School of Information Science and Engineering,Changzhou University,Changzhou 213164,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2020年第5期1098-1105,共8页
Journal of Chinese Computer Systems
基金
江苏省研究生科研与实践创新计划项目(KYCX19_1895)资助.
关键词
拉普拉斯矩阵
费舍尔判别
混合字典学习
人脸鉴别
laplacian matrix
fisher discrimination criterion
hybrid dictionary learning
face recognition
