高志荣,熊承义.基于两级非负线性编码表示的人脸识别[J].中南民族大学学报自然科学版,2014,(4):94-98
基于两级非负线性编码表示的人脸识别
Face Recognition Based on Two-Stage NonnegativeLinear Coding Representation
  
DOI:
中文关键词: 人脸识别  大规模  M-最近邻  线性编码表示  非负系数
英文关键词: face recognition  large scale  M-nearest neighbors  linear coding representation  non-negative coefficient
基金项目:国家自然科学基金资助项目(61471400) ; 湖北省自然科学基金资助项目( 2013CFC118);中央高校基本科研业 务费专项(CZW14018)
作者单位
高志荣1,熊承义2 1 中南民族大学计算机科学学院武汉430074; 2 中南民族大学电子信息工程学院武汉430074 
摘要点击次数: 508
全文下载次数: 705
中文摘要:
      针对大规模人脸识别问题,提出了一种基于两级非负线性编码表示的人脸识别方法.首先利用第一级的线性编码表示,通过在初始的大规模人脸库中寻找对应测试图像的M 最近邻,以消除干扰训练样本并降低训练样本集的规模; 然后以此M 最近邻为训练样本集,通过第二级的线性编码表示实现对测试样本的分类判别.在线性编码表示中,通过进一步引入非负系数约束,更好地改善了分类识别性能.基于AR、ORL 和Yale B 人脸库的实验结果初步验证了文中所提方法的有效性.
英文摘要:
      Aiming to addressing the issue of large scale face recognition,a new two-stage nonnegative linear coding representation based face recognition method was explored in this paper. In the first linear coding representation stage,the scale of training samples set is decreased efficiently by removing the disturbance training samples,which is achieved by finding the M-nearest neighbors from the original large-scale face database. Then based on the new training samples set formed by the M-nearest neighbors,the classification for the probe sample is carried out in the following linear coding representation stage. Additionally,non-negative coefficient constraint is included in linear coding representation to promote the classification performance. Experimental results based on the AR,ORL and Yale B face database demonstrate the effectiveness of the proposed method.
查看全文   查看/发表评论  下载PDF阅读器
关闭