Research On Facial Expression Recognition Method Based On Deep Learning

In recent years,computer vision has developed rapidly in the field of artificial intelligence.How to realize the computer to better understand human emotion and further change the relationship between human and computer has attracted more and more researchers’ attention.Expression recognition is an interdisciplinary subject spanning the fields of artificial intelligence,neurology,computer science,etc.It has great application value in the fields of computer vision,clinical medicine,virtual reality and vehicles,and has greatly promoted the development of science and social progress.It is widely used in social life,and the specific application scenarios include computer-interaction,online education,medical services,etc.In the process of facial expression recognition,face detection is the premise,image preprocessing is the foundation,expression feature extraction is the key,and expression classification is the goal.Effective expression feature extraction can improve the accuracy of expression classification.The main research object of this paper is the feature extraction and expression classification of images,using a combination of traditional algorithms and deep learning algorithms for facial expression recognition.The main research contents of this paper include the following aspects:1.Duo to traditional facial expression recognition methods have complex feature extraction process and cannot extract high-level features.To solve the above problems,a multi-channel fusion expression recognition method is proposed in this paper.Firstly,cascade classifier based on Haar features is used to detect face.Secondly,the local binary pattern LBP is used to extract local texture features.Thirdly,edge detection is performed based on Canny.Fourthly,the obtained face image,LBP texture feature image and edge detection canny image are fused,and the fused image is input into the constructed lightweight neural network for training and recognition.Experiments are carried out on the public image databases Facial Expression Recognition 2013(Fer2013)and Extended Cohn-Kanade(CK+)by using hold-out cross validation method.To a certain extent,the accuracy and robustness of facial expression recognition are improved.2.In order to solve the problems of weak robustness and weak generalization ability of network models,this paper proposed a facial expression recognition method based on improved depthwise separable convolutional network.Firstly,build a neural network model based on the Xception model;Secondly,the inverted residual structure with linear bottleneck was introduced into the network structure;Finally,the extracted features are classified by Softmax classifier,and the entire network model uses Re LU6 as the nonlinear activation function.In the public image database Fer2013 and CK+,the database is divided by 10-fold cross validation for experiments.This method effectively excavates the deeper and more abstract features of the image,and reduces the influence of illumination,posture,etc.Not only the recognition rate of facial expression images is improved compared with the single network model,but also the network model has strong robustness and generalization ability.