基于改进卷积神经网络的人脸表情识别方法研究  

Research on Face Face Recognition Based on Improving Convolutional Neural Network

作　　者：刘明星 LIU Mingxing(Zhengzhou University of Industry Technology,Zhengzhou Henan 450000;Henan Suxing Technology Co.,Ltd.,Zhengzhou Henan 450000)

机构地区：[1]郑州工业应用技术学院,河南郑州450000 [2]河南速兴科技有限公司,河南郑州450000

出　　处：《软件》2025年第2期141-143,共3页Software

基　　金：教育部产学合作协同育人项目“应用型高校计算机科学与技术专业信息化教学模式的探索与实践”(241200367011644)阶段性研究成果。

摘　　要：随着信息技术的不断发展,人脸表情识别技术在辅助医疗、远程教育、智慧交通、视频推荐等多个领域的应用越来越广泛。卷积神经网络集合了深度学习与人工神经网络技术的优势,将其应用于人脸表情识别系统,能够省略传统人脸表情识别技术手工提取面部特征的步骤。该技术对图像进行预处理后利用卷积层提取特征,并利用Softmax分类器进行识别分类。然而,实际应用中,人脸表情图像会受到光照强度、姿势变化、面部遮挡等因素的干扰,从而影响人脸表情的识别效率。本研究提出一种改变LeNet结构的策略,对人脸表情识别技术进行优化。With the continuous development of information technology,facial expression recognition technology is becoming increasingly widely used in various fields such as assisted healthcare,distance education,smart transportation,and video recommendation.Convolutional neural networks combine the advantages of deep learning and artificial neural network technology,and when applied to facial expression recognition systems,they can eliminate the step of manually extracting facial features in traditional facial expression recognition techniques.This technology preprocesses the image and extracts features using convolutional layers,and uses a Softmax classifier for recognition and classification.However,in practical applications,facial expression images are subject to interference from factors such as lighting intensity,posture changes,and facial occlusion,which can affect the recognition efficiency of facial expressions.This study proposes a strategy to optimize facial expression recognition technology by changing the LeNet structure.

关 键 词：卷积神经网络 技术改进 人脸表情识别 

分 类 号：TB183[一般工业技术]