Research And Implementation Of Facial Expression Recognition Algorithm Based On Deep Learning

With the rapid development of deep learning,human facial expression research has become a hot topic.Expression recognition belongs to the field of pattern recognition and has important application value for national defense security,children's education and psychology.For this reason,it has been desired to accurately recognize facial expressions by machines.At the same time,facial expressions,as one of the most powerful social signals,can help to understand each other's inner feelings in daily life communication.However,in the basic expression of human beings,smile is the embodiment of human inner joy,and smile is one of the most common expressions in human communication.Smile detection,as a specific work in basic facial expressions,has enormous potential application value in psychology and human behavior analysis.At the same time,smile detection is especially important for early intervention and diagnosis of children with autism.To solve the above problems,the research on smile detection and seven-class facial expression recognition algorithm for two classifications and the construction of seven types of expression recognition systems will be carried out.The main work of this paper is as follows:(1)Establish a database of children's smile expressions.The database construction process of children's smile expression database is introduced in detail,and the operations of web crawler,manual labeling and image preprocessing are introduced in detail.From the large number of pictures of web crawlers,after professional manual labeling and image preprocessing,2804 is selected.A picture of children's expressions,including 1445 children's smiles and 1359 non-smiles,and finally completed the construction work.(2)Research on children's smile expression recognition.At present,there are very few children's smile expression databases published at home and abroad.Since most smile expression libraries contain adults,adult facial expression features and children's facial features are different and cannot be directly applied to children's smile recognition;therefore,based on VGGNet network,improved BRNet Network,residual network based on migration learning,three kinds of children's smile expression recognition methods.The comparative analysis of the experiment,in which the ResNet network uses the migration learningtechnology to apply to the two-class smile recognition,overcomes the problem of insufficient training database and the disappearance of the gradient as the number of network layers deepens.The ResNet network surpasses traditional methods and classical deep convolutional neural networks in terms of accuracy and performance.(3)The design of the facial expression recognition system.Apply the theoretical research in the thesis to practice,build and optimize six convolutional neural networks,and train the facial expression recognition model on the facial expression database(FER-2013)to identify the seven basic human expressions and build a facial.The expression recognition system,the system presentation interface is simple,interactive and easy to operate.In the field of security,facial expression recognition system is like real-time monitoring of facial expressions.It can monitor dangerous personnel and personnel at important locations.If an emergency occurs,it can quickly respond.