Comparison Among Double-Stream CNN and Local Feature Fusion of CNN in the Facial Expression Recognition Field

In recent years, with the development of deep learning, facial expression recognition has become a hot research direction. In order to improve the accuracy of facial expression recognition, this paper focuses on comparing two improved algorithms based on convolution neural network (CNN), namely, double-stream CNN and CNN local fusion feature facial expression recognition. The dual-stream CNN algorithm uses a variety of advanced technologies for preprocessing and feature extraction; Local feature fusion of CNN is to perform complex processing on local features and finally optimize feature fusion weights. In this paper, the accuracy of face recognition of the two algorithms on CK+ and JAFFE data sets is 98.81% and 96.05%; 94.56%, and 97.08%, respectively. The recognition rate of these two algorithms shows that the optimization effect is very good. By comparing the similarities and differences of the two algorithms, it provides ideas and directions for further improving the accuracy and algorithm.