计算机与数字工程2024，Vol.52Issue(12)：3491-3495,3535,6.DOI:10.3969/j.issn.1672-9722.2024.12.003
基于优化ResNet的人脸表情识别
A Facial Expression Recognition Algorithm Based on Optimized ResNet
摘要
Abstract
Aiming at the problems of low accuracy of traditional facial expression recognition algorithms and many network pa-rameters,a facial expression recognition algorithm based on optimized residual network is proposed.Firstly,two standard convolu-tion layers are used to extract the shallow features of facial expressions.Then,the depth separable convolution hybrid channel atten-tion mechanism is used to improve the residual network to extract the deep features of facial expressions.Finally,softmax function is used to classify the extracted features.Experiments on the public dataset FER2013 and CK+for facial expression recognition show that the classification accuracy is 70.57%and 99.28%respectively.Experimental results show that the algorithm performs well,the network has strong generalization ability,and can play a good role in facial expression recognition in complex situations.关键词
人脸表情识别/轻量型网络/深度可分离卷积/注意力机制/残差网络Key words
facial expression recognition/lightweight network/deep separable convolution/attention mechanism/ResNet分类
信息技术与安全科学引用本文复制引用
徐子凡,程科,袁雪梅,姜元昊..基于优化ResNet的人脸表情识别[J].计算机与数字工程,2024,52(12):3491-3495,3535,6.基金项目
国家自然科学基金项目(编号:61976241) （编号:61976241）
镇江市国际合作计划项目(编号:GJ2021008) （编号:GJ2021008）
江苏省研究生创新训练项目(编号:KYCX21_3487)资助. （编号:KYCX21_3487）
