Research And Application Of Fatigue Detection Algorithm Based On Deep Learning And Micro Expression

In order to improve the quality of teaching,we need to judge the students’ classroom state,find out whether the students are tired in class as soon as possible,and teachers can timely remind the relevant students,make corresponding adjustments to the classroom atmosphere,and improve the students’ classroom efficiency.It is of great significance to study the methods of students’ classroom fatigue detection and apply them to classroom teaching to assist teachers in teaching,so as to provide new ideas for the realization of smart classroom.However,the traditional convolutional neural network can not guarantee the real-time and accuracy of detection at the same time.Therefore,an improved multi task cascaded neural network（mtcnn）is proposed to realize fatigue detection.Firstly,mtcnn is used to locate the key points of human face.Secondly,an accurate eye location method based on multi task constrained learning is proposed.The conventional activation function relu is replaced by leaky relu to avoid the effect of neuron inactivation.When a person is in a state of fatigue,it will be reflected through the micro facial expression.Especially the slight changes of the eyes and mouth can reflect this phenomenon.To solve this problem,this paper proposes a micro expression fatigue detection algorithm based on deep learning.By analyzing the changes of facial micro expressions of relevant personnel,the eye and mouth data sets are constructed,and the eye and mouth state classification model training is completed.Finally,using the trained model,combined with the corresponding judgment standard to achieve fatigue detection.In view of the inaccuracy of eye positioning,this paper takes whether to wear glasses and head posture as auxiliary tasks to carry out accurate eye positioning in the way of multi task learning.Due to the lack of relevant data sets,in order to solve the problem of insufficient and unbalanced data sets,eye and mouth data sets are collected from ZJU data set and yawdd data set respectively,and the data set is expanded by data enhancement.In the eye state and mouth state recognition,considering the small image size of the target area,in order to save time and cost,a more portable convolutional neural network is designed.And through experiments to study the influence of different parameters on the network,and ultimately determine the appropriate network structure and parameters.Finally,aiming at the deficiency of single feature of traditional fatigue detection,a new fatigue detection index is proposed to detect the micro expression,that is,the eye condition when fatigue and the mouth condition when yawning.A new judgment method is proposed by fusing the two fatigue indexes,which makes the fatigue detection results more accurate.This paper studies the fatigue detection method based on deep learning and micro expression,and improves the traditional method.The experimental results show that the fatigue detection method based on deep learning and micro expression has good recognition effect,and the real-time performance has been greatly improved.It has a different interpretation of students’ class state,realizes intelligent classroom teaching,and makes students’ classroom life more colorful.