Research On Expression Classification Method Based On Probability Graph Model

In recent years,deep learning has developed rapidly and has attracted widespread attention at home and abroad.Due to its strong learning ability,deep learning has been successfully applied to multiple pattern classification problems.In real life,the characteristic dimension of the sample is often high,and the acquisition of the sample is always difficult,so the small sample problem is ubiquitous.It is important to study how to improve the accuracy of learning algorithms in small sample sets.This paper proposes a classification method based on probability graph model.First,we give the definition of the aggregation space and the feature space,and prove that the aggregation space can represent the edge probability of the object set,and the feature space can represent the joint probability of the object set.Using the probability graph model theory,the aggregate space and feature space are constructed as node sets to construct weights on the aggregate space set and its own connected edges,and the parameters are updated by learning the training samples.Implement the classification function of the network.In order to verify the validity of the method and the accuracy of classification in small sample sets,validation experiments were performed on 15 Scence and Caltech101 image datasets.The experimental results show that the proposed method has a good classification effect on small sample sets.Secondly,an expression classification method based on probability graph model is proposed.In the traditional expression recognition method research,the researchers used the entire facial expression image as the feature information,thus neglecting the relationship between the various regions of the face.Based on this,this paper proposes an expression region segmentation method,which divides the facial expression image into five face regions.Based on the theoretical basis of the probability map model,five expression classification sub-networks are constructed,and the Softmax classifier is established.And the Softmax classifier forms an expression classification model based on the probability map model to realize the classification of the expression image.Finally,through the experimental analysis on JAFFE facial expression database,CK expression database and Fer2013 expression database,the recognition accuracy of this algorithm in JAFFE facial expression database and CK expression database is 97.78% and 98.95% respectively.Increased accuracy by 1.85% and 5.92%,respectively.The experimental results show that the proposed method has important significance for the improvement of expression classification and recognition structure,and the method of this paper effectively improves the ability to analyze and understand small sample images.The paper has 49 pictures,10 tables and 50 references.