基于改进VGG网络的人脸表情识别  

Face expression recognition based on improved VGG network

作　　者：杨旭 黄昂 肖梦迪 王兴宏 朱琪 YANG Xu;HUANG Ang;XIAO Mengdi;WANG Xinghong;ZHU Qi(Fuyang Preschool Teachers College,Fuyang,Anhui 236000,China)

机构地区：[1]阜阳幼儿师范高等专科学校,安徽阜阳236000

出　　处：《计算机应用文摘》2024年第19期94-96,共3页

基　　金：2023年度安徽省级质量工程项目:大数据服务十大新兴产业特色专业(2023sdxx255);2022年度安徽省级质量工程项目:大数据技术专业教学创新团队(2022cxtd108);2023年度安徽省级质量工程项目:现代教育技术教学创新团队(2023cxtd217)。

摘　　要：在特征提取过程中,传统卷积神经网络常采用单一尺寸的卷积核。随着网络结构的加深,浅层特征无法直接传递给分类器,因此难以准确完整地表达出数据的特征信息。对此,文章提出了基于改进VGG网络的人脸表情识别算法,在VGG16网络中引用Inception和跳跃连接来提高特征的提取和传递能力,从而提高表情识别效果。实验结果表明,改进后的VGG结构在Fer2013数据集上具有比原始VGG网络更高的识别率。In the process of feature extraction,traditional convolutional neural networks often use a single size convolution kernel.As the network structure deepens,shallow features cannot be directly transmitted to the classifier,making it difficult to accurately and completely express the feature information of the data.In response,the article proposes a facial expression recognition algorithm based on an improved VGG network,which uses Inception and skip connections in the VGG16 network to enhance feature extraction and transmission capabilities,thereby improving expression recognition performance.The experimental results show that the improved VGG structure has a higher recognition rate than the original VGG network on the Fer2013 dataset.

关 键 词：VGG16 INCEPTION 跳跃连接 批量归一化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]