Research On Facial Expression Recognition Based On Convolutional Neural Network

Facial expressions are subtle signals in communication,and understanding facial expressions is an important part of understanding communication.With the rise of artificial intelligence,facial expression recognition has attracted more and more attention.Traditional facial expression recognition methods are particularly vulnerable to environmental factors such as illumination and noise,resulting in unstable facial expression recognition.The facial expression recognition method proposed in this paper is based on the convolutional neural network in deep learning.By optimizing the recognition scheme and improving the convolutional neural network,the accuracy of facial expression recognition is not only improved,but also can be stabilized for facial expression recognition.The main research contents include the following points:(1)Although there are many and various facial expression data sets at present,most of the facial expression data sets are filmed from a camera at a certain Angle,so the trained model has certain uncertainty,weak generalization ability to random new data and low robustness.Based on the idea of ROI,this paper made 8 kinds of processing on the facial expression data set,rebuilt a set of its own data set,paid attention to the detailed features of facial expression images,and improved the generalization ability of facial expression recognition model.(2)Traditional lenet-5 convolutional neural network is used for handwritten number recognition.When feature extraction is conducted,low-level details are not taken into account.In this way,if the network gets deeper and deeper,gradient disappearance or explosion problems are likely to occur.In this paper,the facial expression recognition algorithm based on lenet-5 neural network with cross-layer connection of region of interest is proposed,and lenet-5 neural network is improved by using cross-layer connection method,and the features of the low-level network are also taken into account,which not only improves the accuracy of facial expression recognition,but also enhances the robustness of the training model.(3)Generative antagonistic network(GAN)has attracted wide attention due to its good performance in generating high-quality data.GAN can generate synthetic samples simulating real sample distribution from training data,especially GAN has been successfully applied to some face-related tasks.In this paper,a new combined deep learning method for facial expression synthesis and recognition is proposed,which further improves the effect of facial expression recognition.