Research on Facial Expression Recognition Based on Neural Network

For facial expression recognition, this paper proposes a cross-connected AlexNet improved convolutional neural network model. In general, due to lack of image information and noise interference, traditional machine learning methods lack robustness and poor recognition rate. Based on the advantages of deep learning in feature extraction, this paper adds a convolution layer and a pooling layer to the original AlexNet network structure, and uses the cross-connect method to extract low-level features from the network structure and combines advanced functions to construct a Classifier. The trainable features are used to extract feature convolution kernels, and then the pooling layer performs downsampling to eliminate unimportant information, reduce the amount of information calculation, and finally use the Softmax classifier for classification and recognition. Conduct comparative tests before and after the improvement to verify the performance improvement of the improved method.