Research On Facial Expression Recognition Based On Deep Convolutional Neural Networks

Human facial expression can reflect people's emotions and mental activities most intuitively,and it is an important medium for the expression of human's emotions.What's more,it also contains abundant human behavior information and emotion information.It has great application prospects in analysis and recognizing facial expression quickly and accurately in many fields.In the past decade,analyzing and recognizing human's facial expression has become one of the most popular research topics in the field of computer vision.Many works of existing research have been applied in many fields successfully,such as human-computer interaction,character animation,self-driving,and social robots.However,it makes a great difference in the ways of expressing emotions,which leads to different facial expressions.The same expression may show different emotions if the person stay in different environments.If different features between different people(such as gender,skin color,ethnicity,etc.)are taken into account,then these external differences that have nothing to do with facial expression recognition will be amplified.In addition,some different facial expression may share very similar characteristics.There is only slight local differences between this similar expressions and hard to be distinguished.In order to solve these problems,the main work of the paper are as follows:(1)In order to eliminate the influence of different factors such as facial posture,image background and illumination on facial expression recognition,we propose a method to accurately segment the facial expression regions.(2)In order to improve the ability to distinguish between similar facial expressions,we combines the global expression feature and the local expression feature as a new expression feature,and a parallel convolutional neural networks is designed for it.This method can improve the performance of facial expression recognition significantly,achieving an average recognition rate of 94.65%on the CK+ database.(3)A face recognition algorithm based on deep learning is combined with parallel CNN to implement an automatic facial expression recognition system,The system can simultaneously detect and recognize the facial expressions of multiple faces.