Face Emotion Recognition Based On Convolutional Neural Network

Under the influence of the rapid development of artificial intelligence,the cause of intelligent education is booming,and people are paying more and more attention to emotional communication in the process of online education.Face emotion recognition in emotional communication has become an important issue in online education.Aiming at the problem of emotional communication in human-computer interaction,this paper builds a facial emotion recognition system based on convolutional neural network.The main research contents are as follows:1.In view of the problem of too many invalid features in the feature extraction process of face emotion recognition,this paper adds an attention mechanism module CBAM to the VGGNet part of the emotion recognition system,which calculates attention features from both the channel and the space.The figure enhances the original network's attention to target features,making the extracted feature information ignore the unimportant features and making the target features more detailed and comprehensive.2.Aiming at the problem of single model convolutional neural network's insufficient extraction of small facial expression features,multiple parallel structured convolutional neural networks are integrated,and the degree of use of each model's feature is reasonably selected.In this paper,the features of VGGNet and CliqueNet are fused into VCNet in parallel.The feature information contained in the fused features is more comprehensive and representative,the feature representation ability is stronger,and the classification results are more accurate.3.Aiming at the problem of lack of emotion in intelligent education,an emotion recognition system based on convolutional neural network is proposed.The classifier of this system is based on VCNet,and the fused feature information is input into Softmax classifier for feature classification to predict emotion category.The algorithm before and after the improvement is simulated on the expression database JAFFE.The experimental results show that the classification effect of the VCNet proposed in this paper is better than the classification effect of the algorithm before the improvement,which proves the effectiveness of the improved algorithm in facial emotion recognition.The accuracy of the trained classifier was tested,and the average accuracy reached the expected goal.