Research on Facial Expression Recognition Based on LBP and DeepLearning

Traditional algorithm of LBP fails to take the role of the central pixel and relationship between it and the neighborhood pixel into consideration in comparing the gray value of the central pixel and that of the neighborhood pixel, which causes part facial features lost. This paper improves the algorithm of LBP by targeting at the problems above mentioned and puts forward a multidirectional feature of LBP. Through the redefinition of the calculation of the gray value of the central pixel and the neighborhood pixel, this paper obtains a new facial textural feature and designs a seven-layer convolutional neural networks (CNN) for the classification of the features so as to complete facial expression recognition. The experiment carries out verification with JAFFE and obtains the recognition rate of 86.7%, which proves the effectiveness and feasibility of this method.