Facial Expression Recognition Method Based On Deep Learning

With the important value of deep learning technology in different fields,more and more scholars have begun to use deep neural networks to adaptively learn and recognize the features of facial expressions.However,the study of expression recognition is still challenging.First of all,most of the proposed facial expression recognition methods based on deep learning are aimed at the expression of positive or near positive faces,so the recognition performance of non-positive facial expressions is poor.Secondly,deep neural network is extracted in features.The two-dimensional image is often processed directly,and the extracted features may lack information on the facial expression.In view of the above problems,this paper studies the facial expression recognition algorithm based on deep learning.With the problem of the most existing facial expression recognition methods do not study non-positive facial expressions,this paper proposes an expression recognition method based on multi-region learning.Firstly,the face image is divided into multiple different regions by meshing.Then,the convolutional neural network is used to adaptively learn and classify multiple regions,and different weights are assigned to each region according to the recognition result.Finally,the identified results are fused according to their respective weight parameters,and the final recognition result is output.Based on the data samples of CK+ and RAF-DB,the proposed method shows good recognition performance.Facial expression recognition methods based on deep learning usually extract features directly from two-dimensional images.However,these features may not be able to contain information about subtle changes in facial expressions.Therefore,this paper proposes an expression recognition method based on multi-feature fusion.Firstly,the detection algorithm is used to locate face feature points,and face images and geometric feature factor images are obtained.Then,the pixel features and geometric features of face image and geometric feature factor image are extracted and identified by convolutional neural network.Finally,the results were weighted and fused as the final expression classification results.The validity of the proposed method is verified on CK+ dataset,and it has a better effect than the direct learning method of pixel features.