计算机工程 ›› 2011, Vol. 37 ›› Issue (18): 10-11.doi: 10.3969/j.issn.1000-3428.2011.18.004

• 博士论文 • 上一篇    下一篇

采用压缩感知的人脸识别算法

魏冬梅 1,2,周卫东 1   

  1. (1. 山东大学信息科学与工程学院,济南 250100;2. 山东师范大学传播学院,济南 250014)
  • 收稿日期:2011-02-21 出版日期:2011-09-20 发布日期:2011-09-20
  • 作者简介:魏冬梅(1978-),女,讲师、博士研究生,主研方向:图像处理,模式识别;周卫东,教授、博士、博士生导师
  • 基金项目:

    国家自然科学基金资助项目(30870666);山东大学自主创新基金资助项目(2009JC004)

Face Recognition Algorithm Using Compressive Sensing

WEI Dong-mei 1,2, ZHOU Wei-dong 1   

  1. (1. School of Information Science and Engineering, Shandong University, Jinan 250100, China;2. School of Transmission, Shandong Normal University, Jinan 250014, China)
  • Received:2011-02-21 Online:2011-09-20 Published:2011-09-20

摘要: 介绍压缩感知(CS)理论,并将其应用于人脸识别。运用训练数据构造冗余字典,采用随机分布的规范行矢量高斯矩阵构造感知矩阵,对训练图像和测试图像进行感知。利用正交匹配跟踪算法求最小零范数解,在变换域中用近邻法判断测试数据的类别。实验结果表明,用CS进行人脸识别,能避免特征选取的问题,且识别率高、运算速度快。

关键词: 压缩感知, 表示矩阵, 感知矩阵, 稀疏, 人脸识别

Abstract: This paper introduces the theory of Compressive Sensing(CS), and three main problems and their solutions when using CS for face recognition. The over complete dictionary is formed by using the training set, and the random matrix with Gaussian entries builds the sensing matrix with normal row vectors. In the test stage, the sensing matrix is projected onto the test vector, and the minimum l0-norm solution is computed with Orthogonal Matching Pursuit(OMP) algorithm. The distance between the reconstruction vector and the train vector is employed to determine the class of the test data. Experiment results show the CS promising aspects for face recognition has high accuracy and efficiency.

Key words: Compressive Sensing(CS), representation matrix, sensing matrix, sparsity, face recognition

中图分类号: