Scholarlink

<p>The growing dependency on virtual learning environments, mainly in language education, brings about the difficulty of reliably measuring the students’ engagement and emotional involvement. The work describes a computer vision-based emotion recognition system that is meant for the real-time tracking of student engagement in online English classes. The system consists of a new Intelligent Tasmanian Devil-Driven Scalable Residual Neural Network (ITD-SResNet), which combines swarm-based adaptive optimization with deep residual learning to increase the recognition accuracy and the model’s scalability in different online learning environments. The research team collected the data from the Student Emotion-Driven Engagement in Classrooms project, which features eight basic emotions: anger, contempt, disgust, fear, happiness, neutrality, sadness, and surprise. The data was recorded during virtual English classroom sessions that involved continuous video streaming of the students’ faces throughout the lectures and interactions. Preprocessing consisted of two main stages that were very important: face alignment, which optimized the position of the faces in all the pictures to be the same, and histogram equalization, which made the lighting in the pictures similar to each other and thus enhanced the contrast and visibility for the entire picture. When it came to feature extraction, the Local Binary Patterns (LBP) were used for encoding very detailed texture features, which are the very basis of emotion classification. The ITD-SResNet was given the preprocessed dataset to learn from in order to classify emotions. It got very high mean accuracy (0.996), accuracy (0.995), precision (0.993), recall (0.987), and an F1-score of (0.994), surpassing even the best conventional models. The system has real-time visual dashboards for teachers, indicating which emotions are declining or increasing and how many students are participating during the session. This system provides instructors with the opportunity to give dynamic feedback and supports pedagogical adaptations like pace adjustment or interactive reinforcement. It is online English learning through an intelligent and responsive mechanism for tracking and improving student engagement based on emotion-driven insights that eventually gets better.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Computer vision enhanced emotion recognition system for real time participation analysis in virtual english classroom

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月22日

查看全文
  • Xianghui Pan

摘要

The growing dependency on virtual learning environments, mainly in language education, brings about the difficulty of reliably measuring the students’ engagement and emotional involvement. The work describes a computer vision-based emotion recognition system that is meant for the real-time tracking of student engagement in online English classes. The system consists of a new Intelligent Tasmanian Devil-Driven Scalable Residual Neural Network (ITD-SResNet), which combines swarm-based adaptive optimization with deep residual learning to increase the recognition accuracy and the model’s scalability in different online learning environments. The research team collected the data from the Student Emotion-Driven Engagement in Classrooms project, which features eight basic emotions: anger, contempt, disgust, fear, happiness, neutrality, sadness, and surprise. The data was recorded during virtual English classroom sessions that involved continuous video streaming of the students’ faces throughout the lectures and interactions. Preprocessing consisted of two main stages that were very important: face alignment, which optimized the position of the faces in all the pictures to be the same, and histogram equalization, which made the lighting in the pictures similar to each other and thus enhanced the contrast and visibility for the entire picture. When it came to feature extraction, the Local Binary Patterns (LBP) were used for encoding very detailed texture features, which are the very basis of emotion classification. The ITD-SResNet was given the preprocessed dataset to learn from in order to classify emotions. It got very high mean accuracy (0.996), accuracy (0.995), precision (0.993), recall (0.987), and an F1-score of (0.994), surpassing even the best conventional models. The system has real-time visual dashboards for teachers, indicating which emotions are declining or increasing and how many students are participating during the session. This system provides instructors with the opportunity to give dynamic feedback and supports pedagogical adaptations like pace adjustment or interactive reinforcement. It is online English learning through an intelligent and responsive mechanism for tracking and improving student engagement based on emotion-driven insights that eventually gets better.