Research On Facial Expression Feature Extraction And Recognition Algorithm Based On Image Recognition

Emotional expression is the most common way of communication in human daily social interaction.The direct reaction of emotion is expression.In the amount of information of emotional expression,expression can reflect more emotional information than voice,action and oral expression.At present,the main problems of facial expression recognition are as follows: the complex and changeable environment of human beings leads to the influence of some unrelated factors such as illumination,angle,occlusion and background;The extraction of facial expression related feature information has always been the focus of facial expression recognition;After image normalization,most of the interference information can be eliminated,but the occluded image can not be solved by image normalization,which is also a key problem to be considered;After the test samples are compared with the training samples,we need to compare and classify them to realize the output of expression recognition.In view of these problems,this paper has made corresponding improvements in image feature information extraction,occlusion image problem solving and image classification.The main contents and innovations of this paper are as follows:1)The LBP algorithm is improved.It makes better use of the fast speed of traditional LBP algorithm to extract feature information,and improves the accuracy of feature information,which is of great significance for the scene that needs to quickly recognize the expression.The traditional LBP algorithm only expresses the LBP coding value through the connection of each pixel region itself.Based on the traditional LBP,this paper introduces an exclusive threshold m in the pixel region,and compresses the original 8-bit binary code to 4bits.2)An effective image inpainting algorithm is proposed,which can solve the problem of image occlusion in expression recognition.Although the accuracy of image restoration can not meet the requirements of expression recognition,there are some improvements in image restoration.3)Support vector machine(SVM)is selected as the classifier,and KNN algorithm is introduced to overcome the shortcomings of SVM,which improves the problem of long training time.4)The facial expression recognition system is designed and implemented.The system includes face sample collection,model loading and expression recognition module.The sample collection can upload the facial expression image by the user,or collect it on site through the camera.After the model is loaded,the expression type of the facial expression image is displayed in the expression recognition module.