Implementation Of Online Learning State Detection System Based On Expression Recognition

The proposal of a new intelligent education model that combines artificial intelligence with education has driven the rapid development of online learning.However,due to the phenomenon of emotional deficiency in distance education,students’ participation in online learning is relatively low.To eliminate the negative impact of emotional deficiency on students’ online learning,this article conducts in-depth research on deep learning related technologies,designs and implements an online learning system based on student expression recognition technology,monitors students’ learning emotional state changes in real-time,and evaluates students’ learning effectiveness through classroom testing analysis.The main research content of this article is as follows:1)Image acquisition and preprocessing.To address the issues of image exposure and blurriness caused by factors such as lighting and noise during image acquisition,corresponding image preprocessing methods are provided to ensure the accuracy of subsequent facial expression recognition and related technologies.Then,data enhancement is used to amplify data sets to solve the overfitting problem caused by too little data during training of some data sets.2)Face detection.Firstly,we will introduce two face detection methods based on Open CV and MTCNN.Then,by comparing the detection results of the two methods,it was found that MTCNN has better detection performance and robustness for facial images in natural environments,and has low requirements for hardware equipment,which can be used for real-time detection.Therefore,the face detection method based on MTCNN is chosen as the face detection model in online learning scenarios.In response to the issues of missed and false detections in MTCNN face detection methods,propose to replace the NonMaximum Suppression(NMS)algorithm with the Soft Non-Maximum Suppression(SoftNMS)algorithm and fine-tune the parameters to improve and optimize the MTCNN model,Improve the accuracy of student face detection and reduce false positives.3)Propose a facial expression recognition method based on an improved lightweight Convolutional Neural Network SE＿mini＿Xception.This paper first proposes a lightweight convolutional neural networks mini＿Xception model to address the issues of large parameter quantities and time-consuming training for most convolutional neural networks(CNN)models;Subsequently,to address the issue of CNN not being able to focus on the most important areas of facial expression recognition,proposed introducing the channel attention mechanism(SE module)into the mini＿Xception model and naming it "SE＿mini＿Xception".During model training,the SE module assigns appropriate weights to image channels,enhances the information of key areas related to students’ facial emotion expression,suppresses useless information,and improves the model’s feature extraction ability for key areas of students’ facial emotion expression.Finally,the effectiveness and superiority of the SE＿mini＿Xception model were verified through comparative analysis of experimental results.4)Divide students’ classroom emotional states(understanding,questioning,listening,resisting,disdaining)based on their facial expressions,and construct an online learning state detection system based on facial expression recognition technology.This system mainly Utilizing expression recognition model to recognize students’ expressions,providing data support for classroom emotional state judgment.Finally,multiple students were invited to participate in experiments related to online learning video recording,and some facial expression data and test scores were randomly selected for analysis.The results showed that the system can provide accurate data for teachers to understand students’ online learning classroom status.