Facial Expression Classification and Recognition Based on Improved Hybrid CNN-ELM Model

Abstract

In order to further improve the classification accuracy and computational speed of facial expression recognition, this paper proposes an improved facial expression classification and the recognition algorithm based on the hybrid CNN-ELM model. This model uses convolutional neural network (CNN) to learn convolution features of facial expressions, and feeds them to the extreme learning machine (ELM) for face expression classification and recognition. Experimental results show that the model has an accuracy of 91.3% in the JAFFE data set and 89.1% in the fer2013 data set respectively. Compared with CNN algorithm and Gabor feature extraction + ELM algorithm, this model has better test accuracy.