Research On Facial Expression Recognition Method Based On Multiple Features

With the advent of the information age,facial expression recognition technology has been widely used in many fields such as distance education,clinical medicine,and intelligent transportation.However,due to the dark changes of light illumination,the existence of noise and the complex diversity of facial image in practical applications,facial expression recognition still faces many challenges,and many problems need to be solved.This paper studies facial expression recognition using different feature extraction algorithms.The main work of this paper is as follows:1.The LDP algorithm and its improved algorithm are analyzed.To overcome at the defects of LBP operator sensitive to random noise and non-monotonic illumination,a feature extraction algorithm based on DCS-LDP is proposed.The algorithm makes full use of the edge gradient information to compare the response values from the horizontal,vertical and diagonal directions.The multi-order information of the face features on the gradient space is extracted.The JAFFE library and the CK+ expression library are based on noise.The illumination experiment was carried out.The experimental results proved that the DCS-LDP feature extraction algorithm is robust to illumination and noise.It can describe the texture variation between different expressions in different expressions.2.In the practical application of facial expression recognition,the size of the obtained expression images is often inconsistent,and the HOG operator does not have scales and is not deformed,which is not conducive to our good description of facial features.A multi-scale HOG algorithm is proposed on images of different scales combined with an improved HOG algorithm.On the CK+ expression library,we carry out the experiment of pyramid layer and cell scale,and select the optimal parameters of the algorithm,and then compare it with the SIFT feature and the algorithm proposed in [43].The experimental results show that the multi-scale HOG algorithm has better High recognition rate.3.Since a single feature cannot describe the face information from various angles and the ability to use it for expression recognition is limited,this paper proposes a feature extraction algorithm based on multi-feature fusion to fuse LGC features with an improved HOG feature.With texture information and shape information.The comparison experiment between JAFFE library and CK+ expression library shows that the algorithm has better recognition effect than single feature extraction algorithm.