Research On Facial Expression Recognition Based On Convolutional Neural Network

Expression is an indispensable non-verbal communication method in interpersonal communication and plays a role in conveying emotional state and intentions.With the rapid development of artificial intelligence and pattern recognition technology,People's demand for intelligent human-computer interaction is daily on the increase,how to make computer accurately and quickly recognize facial expressions has been widely concerned and studied by many scholars.Deep learning algorithms are widely used in the field of computer recognition by virtue of their powerful feature extraction capabilities.Therefore,research on facial expressions based on deep learning algorithms not only has important theoretical significance,but also has great practical value.By researching and summarizing existing deep learning theories and network models,this paper starts from improving the structure of the existing deep learning model,solving problems such as the generalization ability of network is weak,the complex structure of the model leads to a large amount of calculation.The main research contents of this paper are as follows:1.Propose a new multi-scale feature fusion convolutional neural network model.First,refer to the design concept of the Inception structure to modify the first convolution module of VGGNet,and construct a multi-scale kernel convolutional layer to extract more expression features.In order to prevent the network from losing a lot of useful facial features after multiple convolution and pooling operations,this paper proposes a multi-scale feature fusion strategy,fusing the feature information extracted by different convolutional layers,making the network can capture information at different scales.In addition,the global average pooling layer is used to replace the fully connected layer to reduce the amount of network parameters,add batch normalization layer and Drop Block strategy to prevent network overfitting.Finally,through the joint use of Island Loss and Softmax Loss as a new loss function,extracting more discriminative expression features.This paper conducts experiments on the public facial expression data sets Fer2013 and CK+,the results showed that the improved network can effectively improve the accuracy of facial expression recognition.2.The traditional neural network does not distinguish the effects of different regional features of the face image on expression recognition,but treats the entire face image indiscriminately,this will cause the network extract a large amount of redundant information and noise,which affects the accuracy of expression recognition and cause a waste of network performance.In this paper,the attention mechanism is introduced into the multi-scale feature fusion network,and construct a multi-attention mechanism convolutional neural network model.By designing a new spatial attention module and channel attention module,making the model adaptively acquires the key areas that affect expression recognition from the spatial and channel dimensions,and assigns higher weight to these areas to reduce the influence of redundant information and noise.The experiment was conducted on the public facial expression datasets Fer2013 and CK+,and verified on the validation set.The results show that the improved network model can better recognize the key information of facial expressions,improve the recognition accuracy of error-prone expressions,and further improve the performance of the model.