文献类型：期刊文献

中文题名：基于多维连续情感识别的在线学习风险预警

英文题名：Online Learning Risk Early Warning Based on Multi-dimensional Continuous Emotion Recognition

作者：霍奕[1]

第一作者：霍奕

机构：[1]北京联合大学师范学院,北京100011

第一机构：北京联合大学师范学院

年份：2024

卷号：8

期号：11

起止页码：130-134

中文期刊名：现代信息科技

外文期刊名：Modern Information Technology

基金：教育部人文社会科学研究项目(23YJE880001)。

语种：中文

中文关键词：面部表情识别;情感计算;智能教学系统;学业风险预警系统

外文关键词：facial expression recognition;Affective Computing;intelligent teaching system;academic risk early warning system

摘要：在线教学的教师由于缺乏与学生的情感交互无法像传统面对面课堂那样及时预测学业成绩以提前进行干预。为此,建立基于情感分析的在线教学学业风险预测方法。首先,通过获取效价—唤醒度—控制性(Valence-Arousal-Dominance,VAD)多维情感参数来获得更全面精细的情感信息。其次,利用正交卷积神经网络进行多维情感参数识别。最后,选用多个经典回归模型进行学业成绩和学业风险预测实验,最终选出最适合预测学术风险的模型。实验结果表明,采用正交化卷积约束的神经网络和未进行约束的模型相比,情感参数预测准确性提升;在预测学术成就上引入VAD情感参数比仅使用认知数据的预测准确度明显提升;ADA_RF_EXP模型在最终成绩预测和失败风险警示方面表现最佳。
Online teaching teachers are unable to predict academic performance in a timely manner and intervene in advance like traditional face-to-face classrooms due to a lack of emotional interaction with students.To this end,it establishes an online teaching academic risk prediction method based on sentiment analysis.Firstly,by obtaining the multidimensional emotional parameters of Valence-Arousal-Dominance(VAD),more comprehensive and detailed emotional information can be obtained.Secondly,it uses orthogonal convolutional neural networks for multi-dimensional emotion parameter recognition.Finally,multiple classic regression models are selected for academic performance and academic risk prediction experiments,and the most suitable model for predicting academic risk is ultimately selected.The experimental results show that compared with the unconstrained model,the neural network with orthogonal convolutional constraints improves the accuracy of emotion parameter prediction.The introduction of VAD emotional parameters in predicting academic achievements significantly improves the accuracy of prediction compared to using only cognitive data.The ADA-RF-EXP model performs the best in predicting final grades and warning of failure risks.