Facial Expression Recognition Based on Improved LeNet-5 CNN

Aiming at the influence of local occlusion on expression recognition, this paper proposes an improved cross-connected multi-layer LeNet-5 Convolutional Neural Network model. Under occlusion conditions, traditional machine learning methods are not robust due to lack of image information and noise interference, and the recognition rate is poor. Based on the advantages of deep learning in feature extraction, this paper adds a convolution layer and a pooling layer based on the LeNet-5 model. The low-level features extracted from the network structure are combined with the high-level features to construct the classifier. The implicit features are extracted by using the trainable convolution kernel, and the extracted implicit features are reduced by the pooling layer. Finally, the Softmax classifier is used for classification and recognition. The contrast experiment between the occlusion determination and the occlusion uncertainty is carried out, and the occlusion robustness of the improved method is verified.