一种改进的卷积神经网络的表情识别算法  

An Improved Facial Expression Recognition Technology Based on Convolutional Neural Network

作　　者：付倩倩 李昂[2,3] FU Qian-qian;LI Ang(Wuhan Institute of Posts and Telecommunications,Wuhan 430074,China;School of Telecommunications&Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003,China;Nanjing University of Science and Technology Zijin College,Nanjing 210023,China)

机构地区：[1]武汉邮电科学研究院,湖北武汉430074 [2]南京邮电大学通信学院,江苏南京210003 [3]南京理工大学紫金学院,江苏南京210023

出　　处：《计算机技术与发展》2020年第11期80-83,共4页Computer Technology and Development

基　　金：江苏省高校自然科学基金面上项目(18KJD510004);江苏省普通高校学术学位研究生科研创新计划项目(KYLX160661)。

摘　　要：随着计算机视觉领域的发展,智能化人机交互技术越来越受人们的重视。作为最直接、最有效的情感识别方式,人脸表情识别现已是人机交互领域研究的一大热点和难点。由于人脸识别容易受到光照、旋转、遮挡等复杂因素的影响,传统的人脸识别方法的准确度会大大减少。为了使机器能够快速准确地感应人脸表情,提出以卷积神经网络(convolutional neural network,CNN)来构建表情识别框架,将传统的人工神经网络和深度学习(deep learning,DL)技术结合起来,利用经典的卷积神经网络模型进行分析。将表情分为愤怒、惊讶、高兴、悲伤、恐惧五大类对不同的性别进行识别与分析。结果表明,与传统的表情识别方法相比,该方法有较好的识别效率和时效性,从而可以大大提高人机交互运用的体验感。With the development of computer vision,intelligent human-computer interaction technology has been paid more and more attention.As the most direct and effective emotion recognition method,facial expression recognition has become a hot and difficult issue in the field of human-computer interaction.Because face recognition is easily affected by complex factors such as illumination,rotation and occlusion,the accuracy of traditional face recognition methods will be greatly reduced.In order to make the machine can rapidly and accurately induction facial expressions,we put forward to use convolution neural network(CNN)to build a facial expression recognition framework.Combining traditional artificial neural network and deep learning(DL),we use the classical convolution neural network model to analyze.Expressions are classified into anger,surprise,happiness,sadness and fear to identify and analyze different genders.The results show that this method has better recognition efficiency and timeliness than the traditional expression recognition method,which can greatly improve the experience of human-computer interaction.

关 键 词：人脸表情识别 智能化人机交互 深度学习 卷积神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]