Research On Student Expression Recognition Technology In Classroom Environment Based On Deep Learning

Expression recognition technology in the classroom has significant application value in the field of education.Teachers can improve their efficacy by modifying their teaching tactics after accurately reading their students’ facial expressions to better grasp their emotional states and learning dispositions.Deep neural networks may only be able to learn a limited amount of facial expression features in a setting with many students and complex backgrounds;this could reduce the precision of facial emotion identification.Consequently,it is vital to develop practical feature extraction techniques in order to build effective algorithms for recognizing facial expressions in classroom scenes.This study examines and develops expression recognition algorithms using semi-supervised learning,the attention mechanism,and other theories and methodologies in an effort to address the aforementioned issues.The following are the article’s primary research findings:(1)Self-made Dataset in Classroom Environment(SDCE)appropriate for researching student facial expressions in classroom settings was developed in response to the issue that there are currently few student facial expression datasets,drawing on traditional facial expression datasets and the research of domestic and international scholars on student classroom expressions.The SDCE dataset,which consists of 1206 pictures and 12654 facial expressions,was created using actual university classroom scenarios.(2)This study introduces a semi-supervised deep learning expression identification method with Squeeze-and-Excitation attention mechanism to address the issue that the expression recognition process needs a large number of labeled data(SSA-SE).The attention mechanism can more effectively reduce irrelevant channels or pixels and emphasize important expression aspects than typical neural networks.Large numbers of unlabeled samples can be used in semi-supervised learning to make up for a shortage of data during model training.In this paper,SSA-SE is verified on RAF-DB dataset and Affect Net dataset,and compared with classical algorithms such as Re Mix Match and Fix Match.The results show that SSA-SE has achieved excellent results in two datasets,which is better than the previous classical expression recognition algorithms.(3)This paper combines the spatial-attention mechanism,the channel-attention mechanism,and the self-attention mechanism to address the issue of the low accuracy of single attention mechanisms in facial expression recognition.It also proposes an expression recognition algorithm based on multiple attention mechanisms(Spatial-attention,Channel-attention and Self-attention Network,SCSNET).The feature extraction network,the channel-attention mechanism,the spatial-attention mechanism,and the self-attention mechanism make up the algorithm’s four main network components.The neural network can focus on useful facial expression features with the help of the feature extraction network,channel attention,and spatial attention mechanisms,and the self-attention mechanism,which can suppress the ambiguous labeling brought on by uncertain samples and enhance the model’s expressive ability.The recognition accuracy of this algorithm achieves 86.83% in the experimental assessment on the open dataset RAF-DB,which is more than 1.36% higher than the conventional approach.The experimental design is carried out for each algorithm in order to confirm the efficacy of the aforementioned research findings,and the experimental analysis is compared with the conventional algorithm.Additionally,the facial expression recognition of students in a classroom setting is made possible by combining the MTCNN face detection model and the aforementioned expression recognition algorithm.We can offer more data support and assistance for teachers,students,and parents by examining the students’ concentration levels and facial expressions in the classroom setting.