系统仿真学报 ›› 2018, Vol. 30 ›› Issue (1): 28-36.doi: 10.16182/j.issn1004731x.joss.201801004

• 仿真建模理论与方法 • 上一篇    下一篇

基于改进的字典学习与稀疏表示的人脸表情识别

黎明1,2, 彭秀姣1, 王艳2   

  1. 1.南昌航空大学信息工程学院,江西 南昌330063;
    2.南京航空航天大学自动化学院,江苏 南京210016
  • 收稿日期:2015-12-02 发布日期:2019-01-02
  • 作者简介:黎明(1965-),男,江西南昌,博士,教授,研究方向为智能计算,图像处理与模式识别;彭秀姣(1990-),女,江西九江,硕士生,研究方向为图像处理与模式识别。
  • 基金资助:
    国家自然科学基金(61262019)

Facial Expression Recognition Based on Improved Dictionary Learning and Sparse Representation

Li Ming1,2, Peng Xiujiao1, Wang Yan2   

  1. 1.School of Information Engineering, Nanchang Hangkong University, Nanchang 330063, China;
    2.College of Automation Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China
  • Received:2015-12-02 Published:2019-01-02

摘要: 为克服人脸表情图像识别过程中光照、遮挡等带来的影响,减少稀疏表示分类的时间,提出一种融合HOG特征和改进KC-FDDL(K-means Cluster and Fisher Discrimination Dictionary Learning)字典学习稀疏表示的人脸表情识别算法。对归一化后的表情图像提取HOG特征构成训练集,对训练集进行改进的K-均值聚类的Fisher判别字典学习,利用残差加权的稀疏表示进行表情分类。Cohn-Kanade数据库上的实验结果表明,该算法相比其他的人脸表情分类方法具有耗时低、相似表情分类更准确的优势。

关键词: 遮挡, HOG特征, KC-FDDL字典学习, 稀疏表示, 残差加权

Abstract: In order to overcome the problems induced by illumination andocclude in facial expression recognition and reduce the time required by sparse representation classification, the facial expression recognition algorithm withfusion of HOG feature and improved KC-FDDL dictionary learning sparserepresentation is put forward. Improved K-means cluster and Fisher discriminationdictionary learningis implemented on thetraining setgenerated by extracting HOG features of normalized expression image.Facial expression classification is conducted by thesparse representation with weighted residuals. Experimental results on the Cohn-Kanade databaseshow that this method is lower time-consumingand more accurate for similar facial expression classification than other facial expression classification methods.

Key words: occlude, HOG features, KC-FDDLdictionarylearning, sparse representation, weighted residuals

中图分类号: