Research On Feature Extraction And Recognition Method Of Facial Expression Recognition

Facial expression is a natural and timely reflection of human emotions and plays a very important role in interpersonal communication.Facial expression recognition is often used to analyze the human emotion state and intentions.It has very important applications in the areas of emotion simulation,human psychology,computer vision and humancomputer interaction.Therefore,it has far-reaching research value and practicality which researches on facial expression recognition.The overall framework of facial expression recognition system is designed in this thesis.First,the adaboost algorithm is used to accurately detect and locate the face.Then the obtained image is normalized.And then the features of the image are extracted.Finally,the feature vector is classified.This thesis focuses on the feature extraction and classification recognition methods of facial expression recognition,and proposes improved algorithms.For the study of expression feature extraction,a feature extraction method based on divided local directional pattern(DLDP)is proposed for the problem of the slow extraction speed of LDP feature extraction.First,Kirsch masks in eight different orientations are divided into two sub-directional masks.Then,the edge response values are calculated respectively to obtain DLDP1 and DLDP2.Third,DLDP1 and DLDP2 are concatenated into a single DLDP descriptor.In order to enhance the local features to get more effective information,a new fusion method of DLDP and sobel named sobel-DLDP is presented for feature extraction.The experimental results show that compared with other feature extraction algorithms based on local texture,sobel-DLDP can not only reduce the computation time,but also improve the rate of facial expression recognition.For the research of facial expression classification and recognition,a method of facial expression recognition based on SRC and membership degree analysis is proposed for the problems that HMM,SVM and SRC classification method are not accurate enough for people's mixed expression analysis in real life.Firstly,the components of the tested expressions in various basic expressions are classified by using SRC,and then the proportions of various basic expressions are calculated using the membership degree method.Finally,an effective facial expression recognition analysis is implemented using the principle of maximum membership degree.The experimental results show that the expression recognition method based on SRC and membership analysis can reduce the false recognition rate and improve the accuracy of expression classification.Finally,the system implementation of facial expression recognition is completed through the service robot.A control system of service robot based on expression recognition is constructed,and the recognition result of the expression recognition system is converted into a corresponding control commands to control the motion of service robot in real time.Through repeated experiments,it is verified that the new facial expression recognition method proposed in this thesis not only can effectively control the service robot,but also has strong robustness.