Research Of Face Expression Recognition Based On Graph Convolution Neural Network

Facial expressions play an important role both in the communication between people and in the communication between people and machines.With the development of computer technology,more and more researchers have begun to use deep neural networks to identify facial expressions.This paper is based on the graph convolutional neural network to recognise facial expressions.The main work of the papper is as follows:(1)Graph convolution neural network can process data with non-Euclidean structure.In this paper,a graph convolution neural network is constructed,which is used as a classifier for facial expression recognition.A new method of combining fixed and random points is proposed to converting face image to undirected graph.Firstly,face image is transformed into undirected graph by the method of combining fixed and random points.Then,we put the undirected graph into trained GCN and get the result of facial expression.The experimental results show that the recognition rate of facial expression based on the method of combining fixed points with random points is higher than that based on the method of only using fixed points.But,the recognition rate of facial expression on fer2013 database is lower.(2)In order to overcome the shortcomings of the above-mentioned graph convolutional neural network with low accuracy on complex database,we further propose a fusion graph convolutional neural network feature,LBP feature and HOG feature.Firstly,the features of graph convolution neural network,LBP and HOG are extracted from the facail image.Then,the dimensionality of LBP and HOG features is reduced by PCA and the three features are fused.Finally,a shallow neural network is trained to get the recognition results.The experimental results show that this method improves the recognition accuracy on CK+ dataset,reduces the recognition accuracy on JAFFE dataset and significantly improves the recognition accuracy on fer2013 dataset.