基于GMRF和KNN算法的人脸表情识别  

FACIAL EXPRESSION RECOGNITION BASED ON GMRF AND KNN ALGORITHM

作　　者：张立志[1,2] 王冬雪 陈永超 孙华东 韩小为[1,2] Zhang Lizhi;Wang Dongxue;Chen Yongchao;Sun Huadong;Han Xiaowei(School of Computer and Information Engineering,Harbin University of Commerce,Harbin 150028,Heilongjiang,China;Heilongjiang Provincial Key Laboratory of Electronic Commerce and Information Processing,Harbin 150028,Heilongjiang,China)

机构地区：[1]哈尔滨商业大学计算机与信息工程学院,黑龙江哈尔滨150028 [2]黑龙江省电子商务与信息处理重点实验室,黑龙江哈尔滨150028

出　　处：《计算机应用与软件》2020年第10期214-219,共6页Computer Applications and Software

基　　金：黑龙江省自然科学基金项目(F2018020);哈尔滨市科技局优秀学科带头人项目(2017RAXXJ055)。

摘　　要：传统的人脸表情识别方法往往特征维度低且识别率不高。针对这一问题,提出一种基于高斯马尔可夫随机场(GMRF)的多分块方式特征组合的表情识别方法。将表情图像以不同分块方式分为多个子块;针对每种分块方式下的子块,提取GMRF特征;将不同分块方式的GMRF特征进行组合,用K近邻算法(KNN)进行分类。在JAFFE数据集上进行实验,结果表明该方法对人脸表情识别率达到89.8%,证明了其有效性。The traditional facial expression recognition method tends to have a low feature dimension and a low recognition rate.Aiming at this problem,this paper proposes an expression recognition method based on the multi-blocking feature of Gaussian Markov random field(GMRF).The facial expression images were divided into multiple sub-blocks in different block modes;the GMRF features were extracted for each sub-block in the block mode;the GMRF features of different block modes were combined and classified by K-nearest neighbor(KNN).Experiments on the JAFFE dataset show that our method can recognize facial expressions with a rate of 89.8%,and proves its effectiveness.

关 键 词：表情识别 高斯马尔可夫随机场 图像分块 K近邻算法 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]