作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2012, Vol. 38 ›› Issue (14): 171-172. doi: 10.3969/j.issn.1000-3428.2012.14.051

• 人工智能及识别技术 • 上一篇    下一篇

基于最近特征线的二维非参数化判别分析算法

张 旭1,张向群2,赵 伟1,何岩峰1   

  1. (1. 中国人民解放军71426部队,河南 焦作 454000; 2. 许昌学院计算机科学与技术学院,河南 许昌 461000)
  • 收稿日期:2011-09-19 出版日期:2012-07-20 发布日期:2012-07-20
  • 作者简介:张 旭(1975-),男,博士,主研方向:模式识别,机器学习;张向群,硕士;赵 伟、何岩峰,学士

Two-dimensional Nonparametric Discriminant Analysis Algorithm Based on Nearest Feature Line

ZHANG Xu 1, ZHANG Xiang-qun 2, ZHAO Wei 1, HE Yan-feng 1   

  1. (1. Unit 71426 of PLA, Jiaozuo 454000, China;2. School of Computer Science and Technology, Xuchang University, Xuchang 461000, China)
  • Received:2011-09-19 Online:2012-07-20 Published:2012-07-20

摘要: 提出一种基于最近特征线(NFL)的二维非参数化判别分析算法,用于人脸识别等模式分类问题。该算法在子空间学习阶段运用NFL思想计算训练集中各样例的最近特征距离,计算得到低维投影空间,在低维投影空间中进行分类。通过ORL标准人脸数据库进行实验,结果表明该算法的鲁棒性优于传统算法。

关键词: 最近特征线, 二维非参数化判别分析, 子空间学习, ORL数据库

Abstract: A new subspace learning algorithm called Two-dimensional Nonparametric Discriminant Analysis Algorithm Based on Nearest Feature Line (TDNDA-NFL) is proposed for pattern classification, such as face recognition. The proposed algorithm integrates the idea of NFL and two-dimensional nonparametric discriminant algorithm. It computes the nearest feature distance based on the idea of NFL in the subspace learning stage, then it computes the low-dimensional subspace using two-dimensional nonparametric discriminant algorithm. It classifies in the projected space. In experiments the proposed method is evaluated by the ORL databases and computed with several state-of-the-art algorithms. According to the computed results, the proposed method outperformes other algorithms.

Key words: Nearest Feature Line(NFL), Two-dimensional Nonparametric Discriminant Analysis(TDNDA), subspace learning, ORL database

中图分类号: