摘要
微表情是人类在试图掩饰自己情感时所产生的面部细微变化,在测谎、安防、心理学治疗和微表情识别机器人等方面有着非常广泛的应用,因此微表情识别也开始得到重视。从微表情识别的主流的方法:卷积神经网络及其改进、光流法及其改进、局部二值模式及其改进方法进行分析,对现存的几种方法从使用的算法、准确率、各方法的优缺点、各方法的特点等几个角度进行对比总结;阐述微表情识别目前存在的问题,并对未来的发展方向进行展望。
Micro-expression is a subtle change of human face when trying to cover up their emotions.It is widely used in lie-detect,security,psychological therapy,micro-expression recognition robot and so on.Therefore,micro-expression recognition has been paid more attention.This paper analyzes the main methods of micro-expression recognition:convolution neural network and its improvement,optical flow method and its improvement,local binary pattern and its improved methods.The existing methods are compared and summarized from the algorithms,the accuracy,the advantages and disadvantages of each method,and the characteristics of various methods.Finally,the existing problems of micro-expression recognition are discussed,and the future development direction is prospected.
作者
张人
何宁
ZHANG Ren;HE Ning(Beijing Key Laboratory of Information Service Engineering,Beijing Union University,Beijing 100101,China;Smart City College,Beijing Union University,Beijing 100101,China)
出处
《计算机工程与应用》
CSCD
北大核心
2021年第1期38-47,共10页
Computer Engineering and Applications
基金
国家自然科学基金(61872042,61572077)
北京市教委科技重点项目(KZ201911417048)
北京联合大学人才强校优选计划(BPHR2020AZ01,BPHR2020EZ01)
“十三五”时期北京市属高校高水平教授队伍建设支持计划(CIT&TCD201704069)
北京联合大学研究生科研创新资助项目(YZ2020K001)。
关键词
微表情识别
卷积神经网络(CNN)
LBP-TOP算法
光流法
计算机视觉
micro-expression recognition
Convolutional Neural Networks(CNN)
Local Binary Patterns from Three Orthogonal Planes(LBP-TOP)
optical flow
computer vision
