2021 6th International Conference on Intelligent Computing and Signal Processing (ICSP) | 2021
Face Expression Recognition Based on optimized Convolutional Neural Network
Abstract
In this paper, the facial expression recognition based on convolutional neural network is studied in depth. The face expression recognition network based on VGGNet16 is optimized by adopting the optimization strategy of Batch Normalization, Cross-Entropy Loss Function, Stochastic Deactivation, and Data Enhancement Combination, which increase the accuracy of the optimized network model on the expression recognition of the FER2013 data set by 3.909%, and 1.223% higher than the ICML2013 champion method model. Aiming at the analysis of the recognition of the two types of expressions of disgust and sadness, the original VGGNet16 network model cannot efficiently and completely recognize, and it is easy to recognize confusion. The optimized network model has greatly improved its accurate recognition, and the accuracy rate has been significantly improved by nearly 25% and nearly 10%.