摘 要：针对目前人脸表情识别的准确率偏低、训练速度较慢、泛化能力弱等问题，提出了改进的 VGGNet，添加 BN 算法和 PReLU 激活函数，在图像预处理时加入高斯滤波和直方图均衡化，并且使用 FER2013、AffectNet、JAFFE、CK+ 四种数据集进行比较分析。最终的实验结果表明，该模型在四种数据集上的识别准确率都有所提高，在四种数据集上的准确率达到73.52%、84.66%、94.28%、95.26%。在测试集上的泛化能力较强，训练速度也变快。
关键词：卷积神经网络；激活函数；BN 算法；表情识别
DOI:10.19850/j.cnki.2096-4706.2021.23.026
中图分类号：TP391.4；TP183 文献标识码：A 文章编号：2096-4706（2021）23-0100-04
Research on Facial Expression Recognition Based on Improved VGG Model
ZHANG Shibao, WANG Wentao
(Nanjing University of Information Science & Technology, Nanjing 210044, China)
Abstract: In view of the low accuracy, slow training speed, and weak generalization ability of facial expression recognition at present. An improved VGGNet is proposed, adds Batch Normalization (BN) algorithm and PReLU activation function. Gaussian filtering and histogram equalization are used in image preprocessing, and four data sets of FER2013, AffectNet, JAFFE, CK+ are used for comparative analysis. The final experimental results show that the recognition accuracy of the model has been improved on the four data sets, and the accuracy of the mode on the four data set has reached 73.52%, 84.66%, 94.28% and 95.26%. The generalization ability of the model on the test sets is strong, and the training speed is also faster.
Keywords: convolutional neural network; activation function; BN algorithm; facial expression recognition
参考文献：
[1] 何颖，陈淑鑫，王丰 . 基于 HOSVD 分类的非特定人脸表情识别算法 [J]. 计算机仿真，2021，38（10）：193-198.
[2] 张庆，代锐，朱雪莹，等 . 基于链码的人脸表情几何特征提取 [J]. 计算机工程，2012，38（20）：156-159.
[3] LAJEVARDI S M，HUSSAIN Z M. Automatic facial expression recognition：feature extraction and selection [J].Signal,Image and Video Processing，2012，6：159-169.
[4] MINAEE S，MINAEI M，ABDOLRASHIDI A. DeepEmotion：Facial Expression Recognition Using Attentional Convolutional Network [J/OL]. arXiv：1902.01019 [cs.CV].[2021-11- 03].https：//arxiv.org/abs/1902.01019v1.
[5] JIAO Z J，QIAO F C，YAO N M，et al. An Ensemble of VGG Networks for Video-Based Facial Expression Recognition [C]//2018 First Asian Conference on Affective Computing and Intelligent Interaction.Beijing：[s.n.]，2018：1-6.
[6] ZHONG Y X，QIU S H，LUO X S，et al. Facial Expression Recognition Based on Optimized ResNet [C]//2020 2nd World Symposium on Artificial Intelligence（WSAI）.Guangzhou：IEEE， 2020：84-91.
[7] 程换新，王雪，程力，等 . 基于 CNN 和 LSTM 的人脸表情识别模型设计 [J]. 电子测量技术，2021，44（17）：160-164.
[8] SIMONYAN K，ZISSERMAN A. Very Deep Convolutional Networks for Large-Scale Image Recognition [J/OL].arXiv：1409.1556 [cs.CV].[2021-11-03].https://arxiv.org/abs/1409.1556.
[9] LUCEY P，COHN J F，KANADE T，et al. The Extended Cohn-Kanade Dataset (CK+)：A complete dataset for action unit and emotion-specified expression [C]//2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition-Workshops. San Francisco：IEEE，2010：94-101.
[10] GOODFELLOW I J，ERHAN D，CARRIER P L，et al. Challenges in representation learning：A report on three machine learning contests [J] Neural Networks，2015,64：59-63.
[11] MOLLAHOSSEINI A，HASANI B，MAHOOR M H. Affectnet：A Database for Facial Expression，Valence，and Arousal Computing in the Wild [J].IEEE Transactions on Affective Computing， 2019，10（1）：18-31.
[12] LYONS M J，KAMACHI M，GYOBA J. Coding Facial Expressions With Gabor Wavelets （IVC Special Issue） [J/OL].arXiv： 2009.05938 [cs.CV].[2021-11-03].https://arxiv.org/abs/2009.05938.
作者简介：张士豹（1996—），男，汉族，安徽滁州人，助教， 硕士在读，研究方向：图像处理；王文韬（1998—），男，汉族， 江苏苏州人，助教，硕士在读，研究方向：图像处理。