Research And Implementation Of Facial Expression Recognition

Facial expression recognition is an important part of affective computing and human-computer interaction.It is a research hotspot in the field of pattern recognition and artificial intelligence.It has a wide application prospect and potential market value.Generally speaking,Facial expression recognition system is composed of three parts:face image preprocessing,image feature extraction and dimension reduction,and facial expression classification.In this paper,based on the in-depth study of related technologies,according to the characteristics of the facial expression image,considering the global and local characteristics,the paper puts forward a reasonable scheme and a new strategy to eliminate the individual differences,enhance the division,the adjustment of the nearest neighbors and the fusion model.The main work of this paper can be summarized as follows:(1)Pretreatment part:using the JAFFE expression database as the experimental data,before the expression feature extraction,the advanced pedestrian eye detection.Then the image is rotated,cut and normalized.After the image preprocessing step to leave as much as possible useful features to remove irrelevant factors such as hair,ears and different attitudes.(2)Feature extraction part of expression image:?The whole facial region is selected as the global information,and the three sub regions of the eyes,nose and mouth are used as local information;?Using differential image to reduce the influence of individual specificity on facial expression;?The improved feature extraction method based on SLLE is proposed:Firstly,the class information is used to carry out nonlinear reward and punishment for the distance between samples;Secondly,the information entropy is introduced to measure the complexity of the nearest neighbor points of the sample,and then the nearest neighbor spatial scale is adjusted to adaptively select the number of nearest neighbors,and the reliability of the nearest neighbor reconstruction is increased.(3)Expression image feature classification:A two phase test sample sparse representation method is used to classify the facial expression features,and an integrated model is proposed based on the weighted sum of the residuals and the confidence level at the decision level to fuse the global and partial results.Specifically,this paper presents a two phase test sample sparse representation method for face expression recognition.The global face is divided into several different parts,the recognition results are obtained on each component,and then the recognition results of each part and the global recognition results are fused by weighted voting method.The experimental results show that the proposed method can improve the recognition performance compared with the single global pattern recognition.