计算机应用 ›› 2010, Vol. 30 ›› Issue (4): 964-966.

• 模式识别 • 上一篇    下一篇

基于二维多尺度局部Gabor二进制模式特征的表情识别

张铮1,赵政1,袁甜甜2   

  1. 1. 天津大学计算机科学与技术学院
    2.
  • 收稿日期:2009-10-25 修回日期:2009-12-07 发布日期:2010-04-15 出版日期:2010-04-01
  • 通讯作者: 张铮
  • 基金资助:
    一体化网格数据挖掘平台关键问题研究

Expression recognition based on 2D multi-scale block local Gabor binary patterns

  • Received:2009-10-25 Revised:2009-12-07 Online:2010-04-15 Published:2010-04-01
  • Contact: Zheng Zhang

摘要: 为了在独立于个体身份的面部表情识别中取得更加理想的效果,提出了一种基于二维多尺度局部Gabor二进制模式(MB-LGBP)特征的识别方法。对于表情识别而言,MB-LGBP已被证明了是一种局部和整体上都具有很强表征能力的描绘子。将MB-LGBP与灰度共现矩阵(GLCM)结合起来得到了可以更好地描述局部纹理空间结构特性的二维MB-LGBP特征。在识别中,分别选择了支持向量机(SVM)和基于卡方距离的K-最近邻(KNN)分类器,并对结果进行了比较。实验结果证明了二维MB-LGBP特征相比于MB-LGBP以及其他一些主要的表情识别特征的优越性。

关键词: Gabor滤波器, 灰度共现矩阵, 表情识别, 支持向量机, 局部二进制模式

Abstract: In order to accomplish subject-independent facial expression recognition, a facial expression recognition approach based on 2D Multi-scale Block Local Gabor Binary Patterns (2D MB-LGBP) was presented. MB-LGBP features have been proved to be both locally and globally informative for expression recognition. This research combined the idea of MB-LGBP with the concept of Gray Level Co-occurrence Matrix (GLCM) to achieve the 2D MB-LGBP features, which can encode the local textures with structure information. In recognition, SVM classifier was utilized and its performance was compared with the traditional weighted Chi-square distance based paradigm. The experimental result proves the superiority of the 2D MB-LGBP composite features to MB-LGBP and some other popular features in expression recognition.

Key words: Gabor filter, Gray Level Co-occurrence Matrix (GLCM), expression recognition, Support Vector Machine (SVM), Local Binary Pattern (LBP)