基于轻量卷积网络多层特征融合的人脸表情识别  

Facial Expression Recognition by Merging Multilayer Features of Lightweight Convolutional Networks

作　　者：申毫 孟庆浩 刘胤伯[1,2,3] Shen Hao;Meng Qinghao;Liu Yinbo(School of Electrical and Information Engineering,Tianjin University,Tianjin 300072,China;Institute of Robotics and Autonomous Systems,Tianjin University,Tianjin 300072,China;Tianjin Key Laboratory of Process Detection and Control,Tianjin 300072,China)

机构地区：[1]天津大学电气自动化与信息工程学院,天津300072 [2]天津大学机器人与自主系统研究所,天津300072 [3]天津市过程检测与控制重点实验室,天津300072

出　　处：《激光与光电子学进展》2021年第6期140-147,共8页Laser & Optoelectronics Progress

基　　金：国家重点研发计划(2017YFC0306200);国家自然科学基金(61573253)。

摘　　要：基于深度学习的表情识别方法存在参数量大、实时性差等问题,针对此问题,提出一种基于轻量卷积网络的多层特征融合的人脸表情识别方法。首先使用改进的倒置残差网络为基本单元搭建轻量卷积网络模型,然后采用池化、1×1卷积、全局平均池化法筛选卷积网络中的浅层特征,并对这些筛选的浅层特征与深层特征进行融合用于表情识别。在两个常用的真实表情数据集RAF-DB和AffectNet上对所提方法进行测试,识别准确率分别达85.49%和57.70%,且模型参数量仅有0.2×10^(6)。Current methods of expression recognition based on deep learning exhibit problems such as a large number of parameters and poor real-time performance.To resolve these problems,this paper proposes a facial expression recognition method by merging the multilayer features of a lightweight convolutional network.First,the improved inverted residual network was considered as the basic unit for building the lightweight convolutional network model.Then,the shallow features associated with the convolutional network subjected to the methods of pooling,1×1 convolution,and global average pooling were merged with the deep features of the network for expression recognition tasks.The proposed method was verified by employing RAF-DB and AffectNet,which are commonly used expression datasets.Results indicate that the facial expression recognition accuracies of the proposed method when considering the RAF-DB and AffectNet datasets are 85.49%and 57.70%,respectively.Furthermore,the number of model parameters is only 0.2×10^(6).

关 键 词：图像处理 表情识别 卷积神经网络 浅层特征 深层特征 多层特征融合 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]