Research On Facial Expression Recognition Based On Two Improved Convolutional Neural Networks

Facial expression is the most direct and effective emotion mode.At present,facial expression recognition is widely used and is an important part of human-computer interaction.With the rapid development of deep learning,facial expression recognition technology will be more widely used.The existing facial expression recognition technology has been applied in many fields,such as the expression recognition of teachers and students in online teaching and the application of intelligent medical treatment.It can be applied to the diagnosis of mental patients,face brushing and blinking payment everywhere in shopping malls and other aspects of life.Despite such a wide range of applications,facial expression recognition is still a difficult problem in some complex scenes.In the convolution neural network of deep learning,the challenge of large-scale facial expression recognition at present mainly lies in uncertainty,which comes from ambiguous facial expressions,low-quality facial images and the subjectivity of the tagger.In addition,when the current facial expression recognition technology is applied to large-scale facial expression database,the accuracy of facial expression recognition is low,The research of facial expression recognition is still challenging.To solve this problem,this paper focuses on optimizing the structure of convolutional neural network and improving the accuracy of facial expression recognition.The main research work completed in this paper is as follows:1.According to the development process and research direction of facial expression recognition,I learned about the network model structure and related activation function of facial expression recognition based on deep learning,and understood some channel attention mechanisms to capture important key points.After comparing several convolutional neural network structures in the experimental process,ResNet18 and RepVGG-A0 network models are used as the backbone network of facial expression recognition to study,improve and optimize them.Experimental results show that the improved network model improves the accuracy of facial expression recognition.2.In order to optimize the deep learning network and further improve the accuracy of facial expression recognition,this paper proposes an improved ResNet18 network model.The improved network model replaces the ReLu activation function after the convolution layer of the backbone network with the mish activation function.After the feature extraction of facial expression pictures,Softpool is used to improve the information loss caused by down sampling,which is conducive to distinguish similar key points,so as to improve the accuracy of facial expression recognition.The improved network model is verified by experiments on RAF-DB,CK + and JAFFE data sets.The accuracy of human face expression recognition reaches 87.54%,98.95% and 98.99% respectively,which is higher than that of the original ResNet18 network model.It is proved that the improved model is effective for facial expression recognition and has a certain application prospect.3.In the deep learning convolutional neural network,the challenge of large-scale facial expression recognition mainly lies in uncertainty,which comes from ambiguous facial expressions,low-quality facial images and the subjectivity of the tagger.In order to solve this problem,this paper proposes an improved network model based on RepVGG-A0,which introduces an effective channel attention mechanism module,that is,insert the ECANet channel attention mechanism before the ReLu activation function after the convolution layer.The method of re labeling is used to give the uncertain or incorrectly labeled facial expression pictures,re give the false labels,and use the re labeling module to modify the labels of the pictures.Experiments on RAF-DB and FER2013 data sets show that the accuracy of human face expression recognition reaches 88.20% and 75.20% respectively,which is significantly higher than that of the original network.It is proved that the improved model is effective for facial expression recognition.