Research On Facial Expression Recognition Algorithm Based On Local Features

With the increasing development of science and technology,people have developed great interest in the field of human-computer interaction.Facial expressions are the most powerful and natural nonverbal emotional communication method,so that identifying facial expressions can make the human-computer interaction mode more humane.Nowadays,facial expression recognition technology has been widely used in criminal investigation,medical systems,e-commerce and other fields.Although the facial expression recognition technology has made great progress in the past few decades,there are still many challenges.For example,the face image is easily affected by external factors such as light intensity and noise,and the expression itself exists complexity and diversity.Therefore,extracting facial expression features effectively and accurately is the key in facial expression recognition systems.This paper studies different feature extraction algorithms and proposes three improved algorithms,namely EuLBP algorithm,EDLBP algorithm and facial expression recognition algorithm based on improved distance features,to improve the recognition rate of facial expressions and reduce the complexity of the algorithm.The main research work of this paper is as follows:1.This paper proposes an LBP algorithm by using weighted Euclidean distance,and it is named as the EuLBP algorithm.Since LBP algorithm did not consider that the difference of distances between the 8 pixels in the neighborhood and the center pixel,the EuLBP algorithm adopts different weights for neighboring pixels according to the different distances,and can correct the shortcomings of the LBP algorithm,which is susceptible to light interference and single window.On the OuTex,UIUC,and KTH-TIPS databases with illumination diversity and texture rotation changes,the EuLBP algorithm is compared with various improved LBP algorithms.The experimental results show that the EuLBP algorithm can effectively improve the robustness of the original algorithm to changes in illumination and rotation.2.This paper proposes an improved LBP algorithm based on the combination of Euclidean distance and differential coding(ED),which is named as the EDLBP algorithm.The EDLBP algorithm adopts different weights for neighboring pixels according to the different distances.In addition,for pixels with the same distance from the center point,a new type of differential encoding method is used,which is based on the gray value of the previous pixel and the center pixel.On the OuTex,UIUC,and KTH-TIPS databases with illumination diversity and texture rotation changes,the EDLBP algorithm is compared with various improved LBP algorithms.The experimental results show that the EDLBP algorithm can effectively improve the robustness of the original algorithm to changes in illumination and rotation.In particular,experiments conducted on the CUReT database prove that the EDLBP algorithm can be effectively used for texture classification with complex environmental changes.Tests conducted on the FERET database prove that the EDLBP algorithm has great advantages when using the ?~2 distance to measure the similarity of two face images.3.This paper proposes a facial expression recognition algorithm based on improved distance features.The facial expression recognition algorithm based on distance features neglected the high symmetry of the face image,resulting in high algorithm complexity.Therefore,the improved algorithm can improve the original algorithm from the perspective of high symmetry of the human face,and reduces the 210 distance features of the original algorithm to 105,which is a 50%reduction in feature distance.In addition,the improved algorithm is used to perform classification and recognition on the CK+,JAFFE,and MMI databases,and compared with the original algorithm.It demostrated that for these different networks and different facial expression databases,the recognition rate of the improved algorithm is slightly improved compared with the original algorithm.Although the number of required convergence iterations increased,the required convergence time decreased by 30%because the distance features participating in training were reduced by half.Therefore,the facial expression recognition algorithm based on the improved distance feature greatly improves the recognition performance of the facial expression classification system.