Analysis And Research On Students' Classroom Fatigue Based On Deep Learning

With the improvement and optimization of education system and equipment,it is an inevitable trend to introduce fatigue detection technology into students' classroom,how to keep students in an efficient learning state has always been a research hotspot in various fields,especially in the educational circles.In order to ensure the learning efficiency of the students in the classroom,it is very important to find out the fatigue state of the students as soon as possible and remind them in time.Therefore,it is of great value and significance to bring the fatigue detection system into the classroom in society and academia.In order to overcome the problems of low detection rate of facial features and too single index of fatigue judgment in the current mainstream detection technology,this paper trains an efficient learning network combined with CNN,which improves the accuracy of face and face feature detection,and introduces a new variable as the criterion of fatigue judgment on the basis of PERCLOS eye judgment.The mouth'state is another index to judge fatigue.By sampling and analyzing the students' eyes and mouth in class,the fatigue index is compared and the judgment of fatigue is given.The experimental results prove the importance of the self-training network,the effectiveness of introducing new evaluation indicators and the application value of this paper.The main work accomplished in this paper is as follows:(1)The function of face detection and identity recognition based on convolutional neural network in deep learning is realized,and the database is constructed.But in the actual teaching classroom,students will have a large number of body movements to occlude the face area.Therefore,we use two directions to capture the face video stream for students.We choose two video streams with smaller occlusion for analysis.In addition,we add a large number of lateral face occlusion maps with different proportions to the image data set.The maximum occlusion ratio was detected by experiments.The whole image data set contains 1811 complete face pictures,on which other operations are performed.(2)The feature classifier was used to screen out the feature images of eye and mouth regions,including 885 open-eye pictures and 507 closed-eye pictures.514 images with excessive mouth opening and 996 normal images completed the construction of the experimental database.The network model trained by image data sets in complex environments can improve the detection rate and anti-jamming ability of face,and can be used normally in more complex environment(3)A fatigue judgment index is proposed through experiments.The new index introduces new variables in the principle of PERCLOS,and takes mouth opening as one of the indicators of fatigue judgment.The specific judgment method is determined by exploratory experiments,which greatly improves the accuracy of fatigue detection.(4)Complete the training process of self-designed convolutional neural network.The influence of various parameters on the accuracy of the network model was studied through experiments,and the network structure suitable for this experiment was determined.After training,the classification model of eyes opening and closing and the approximate classification model of mouth opening and closing were obtained.(5)The fatigue detection system based on in-depth learning is implemented.Through three convolutional neural network models,the information that can identify the student's identity,judge the opening and closing of the eyes,and judge the opening and closing of the mouth are obtained respectively,and then the new fatigue judgment index is combined.Give the final judgment of whether the students are tired or not,and complete the detection target of the students' fatigue in the classroom.