Design and Experiment of Facial Expression Recognition Method Based on LBP and CNN

Abstract

Aiming at the poor stability of traditional facial expression recognition methods, the feature extraction method is affected by the external environment such as illumination and posture, and an improved convolutional neural network (CNN) model is proposed. A local binary pattern (LBP) image is extracted from the facial expression image, Combine original face image and LBP image as training data set. Firstly, the expression features are implicitly extracted by means of continuous convolution. Then the extracted implicit features are subsampled by the maximum pooling method. Finally, the Softmax classifier is used to classify the facial expressions. The experimental results show that the improved CNN model trained by adding LBP feature information in the dataset has high recognition accuracy and robustness.