Research on Facial Expression Recognition Algorithm Based on CNN

Jianan Meng, Liguo Zheng


In order to realize the recognition of facial expression quickly and accurately, the traditional convolutional neural network is improved in this paper, and the research of facial expression recognition based on LeNet-5 convolutional neural network algorithm is proposed. Firstly, the image is preprocessed, including normalization and geometric transformation. Secondly, the convolution neural network model based on LeNet-5 model is designed, in which the Flatten layer is added, the Dropout strategy is introduced, and the LeakReLU is used as the activation function. At last, the experimental results show that the recognition rate of the improved algorithm is better than that of LBP, sift and hog, but it takes the longest time and the largest amount of computation.



  • There are currently no refbacks.