Facial Expression Recognition Based on Complete Local Binary Pattern and Convolutional Neural Network

Abstract

Facial expression recognition is a challenging problem in computer vision as changes in the background and location of the same person may also lead to difficulties in the recognition. In order to improve the poor effect of facial expression recognition algorithms due to it’s easy to be affected by illumination and posture in real life environment, a facial expression recognition algorithm based on CLBP and convolutional neural network is proposed. The algorithm uses HOG feature and Adaboost algorithm to cut out images with face and eyes, extracts CLBP fusion feature images, constructs a seven-layer convolutional neural network model with two convolutional and pooling layers, two full connection layers and one Softmax classification layer to recognize six expressions of anger, aversion, fear, happiness, sadness and surprise. Four sets of comparative experiments are designed on the public data set CK+, a recognition rate of 98.60% is achieved, better than the existing mainstream methods.