Design Of Student Learning State Detection System Based On Facial Features

In the traditional teaching classroom,teachers can only evaluate the learning process of students through classroom observation and questioning.Due to the limited conditions,it is difficult for teachers to understand the learning status of each student in the classroom.A comprehensive and scientific analysis of each student's learning status enables teachers to develop a more effective teaching plan based on the analysis results.With the help of the analysis,each student's classroom learning efficiency can be improved in a targeted manner.While the change of facial features can be used as an important basis for judging the learning state of students.Since students do not have enough facial features during the learning process,it seems necessary to study the technology of judging students' learning state based on the changes of limited facial features.This thesis studies and analyzes the techniques of judging students' learning state,and focuses on the study of students' learning state according to the changes of eye and mouth characteristics in facial features.Based on this,a student learning state detection system based on facial features is designed.The basic process of the system is to first obtain the student's class video through the camera,then perform image preprocessing,face detection and face recognition,and then locate the feature points of the eyes and mouth in the detected face image,and The state change algorithm of eyes and mouth is designed to calculate the corresponding characteristic parameter values.Finally,the students' learning state is classified and detected through the changes of parameters of eyes and mouth,and the students' learning state is classified into focus,doubt and fatigue.State.The main research contents and work of this thesis include:(1)Face recognition using the improved AlexNet network,and optimizing the convolutional layer and the fully connected layer in the network structure respectively,in order to improve the accuracy and shortening of algorithm identification.The time of face recognition meets the requirements of high real-time performance of the learning state detection system;(2)The AAM algorithm is used to mark and track facial feature points,and the AAM model of the face is constructed to extract the eye and mouth feature points of the students.Using the P80 standard PERCLOS method and the blink frequency to judge the student's fatigue level,combined with the mouth's aspect ratio and mouth angle curvature to jointly determine the student's learning state;(3)the calculated eye aspect ratio,PERCLOS value,blink The parameters of frequency,mouth aspect ratio and mouth radians are normalized.As the feature input vector,the support vector machine classifier is used to classify and recognize the three learning states of the students,and the classification results are detected.(4)A set of easy-to-use learning state detection system was designed and developed.Face detection,face recognition,eye and mouth state detection and learning state detection were performed by related algorithms.In summary,each student's learning state at each time period is obtained,and the corresponding test results are fed back in real time according to the needs of the system operator.Using the system designed in this thesis,16 students under a certain course were tested for learning state.The experimental results show that the morphological changes of eyes and mouth can effectively judge which of the three learning states of the students' learning state belongs to concentration,doubt and fatigue.The accuracy of detection is high and has certain application value.