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 ›› 2019, Vol. 45 ›› Issue (5): 261-266. doi: 10.19678/j.issn.1000-3428.0052002

• 多媒体技术及应用 • 上一篇    下一篇

基于信息增量特征选择的微表情识别方法

张延良,卢冰   

  1. 河南理工大学 物理与电子信息学院,河南 焦作 454150
  • 收稿日期:2018-07-03 出版日期:2019-05-15 发布日期:2019-05-15
  • 作者简介:张延良(1979—),男,副教授、博士,主研方向为微表情识别、人工智能、信号处理、机器学习;卢冰,硕士研究生。
  • 基金资助:

    国家自然科学基金(61571339);网络与交换技术国家重点实验室开放课题(SKLNST-2016-1-02);河南理工大学博士基金(B2012-100)。

Micro-expression recognition method based on information gain feature selection

ZHANG Yanliang,LU Bing   

  1. School of Physics and Electronic Information Engineering,Henan Polytechnic University,Jiaozuo,Henan 454150,China
  • Received:2018-07-03 Online:2019-05-15 Published:2019-05-15

摘要:

基于LBP-TOP、HOG-TOP、HIGO-TOP特征描述子的微表情识别方法通常提取到的特征向量维度较高,计算复杂度较大,运行时间较长,识别准确率较低。为此,提出一种基于信息增量(IG)特征选择的识别方法。运用IG特征选择方法对高维度特征向量进行降维,提高识别效率。运用支持向量机分类器的线性核、卡方核、直方图交叉核进行留一交叉验证,以完成分类任务。在SMIC和CASME2数据集上进行实验,结果表明,经IG选择后,特征向量在2个数据集上的识别准确率分别达到76.22%和73.68%,分类所需时间分别缩短为原方法的3.67%和3.64%,验证了该方法的有效性。

关键词: 微表情识别, 信息增量, 特征描述子, SVM分类器, 核函数

Abstract:

Micro-expression recognition method based on feature descriptor of LBP-TOP,HOG-TOP and HIGO-TOP usually extract feature vectors with high dimensions,and have high computation complexity,long running time and low recognition accuracy.Therefore,a recognition method based on Information Gain(IG) feature selection is proposed.The IG feature selection method is applied to reduce the dimensions of feature vectors and improve the recognition efficiency.The Leave-One-Subject-Out Cross Validation is performed for the micro-expression classification with linear kernel,chi-square kernel and histogram intersection kernel of Support Vector Machine(SVM) classifier.On the SMIC and CASME2 datasets,the recognition accuracy of feature vectors selected by IG achicves 76.22% and 73.68% respectively.And the time required for classification is only 3.67% and 3.64% of the original method.These results prove the effectiveness of the proposed method.

Key words: micro-expression recognition, Information Gain(IG), feature descriptors, SVM classifier, kernel functions

中图分类号: