Facial Expression Recognition Based on Sparse Autoencoder and Shallow Convolutional Neural Network

Abstract

Facial expression recognition is an important research issue in the pattern recognition and accurate recognition is a great challenge, especially in real scenes. In this paper, we propose a method of facial expression recognition based on sparse autoencoder and shallow convolution neural network, which can effectively solve the problems of insufficient feature extraction of shallow convolution neural network and limited sample datasets. The experimental data adopts the ICML2013 facial expression recognition contest’s dataset (FER-2013) and this data is more difficult to identify. The images are preprocessed before the experiment and the key parts are selected as the input of the model. After these features are extracted many times, they are finally classified by softmax classifier. The experimental results indicate that the model performs well on the FER2013 dataset and the accuracy has been greatly improved compared with other methods.