基于深度学习的学生学习情感模型建立与分析  

Establishment and Analysis of Students’learning Emotion Model Based on Deep Learning

作　　者：周江[1] 李锋[1] 蔡臻[1] ZHOU Jiang;LI Feng;CAI Zhen(Guangdong Communication Polytechnic,Guangdong Guangzhou 510650,China)

机构地区：[1]广东交通职业技术学院,广东广州510650

出　　处：《信息与电脑》2023年第2期104-107,共4页Information & Computer

基　　金：2022年广东省科技创新战略专项资金项目(攀登计划)(项目编号:pdjh2022b0857);2020年广东交通职业技术学院校级科研项目(项目编号:GDCP-ZX-2020-006-N1)。

摘　　要：传统的情感模型仅仅关注学生学习表情与对应的学习情感之间的关系,而忽略了不同学习情感之间的关系,因而导致学生学习表情识别准确率相对较低。基于此,建立学生学习表情三维状态空间情感模型,并在其中引入Maxout神经元,从而构建优化的三维状态空间情感模型,进一步解决三维梯度弥散问题,更好地优化系统的训练过程,在本模型中还引入了情感分类器的概念,实现对学生学习表情情感状态的有效分类,从而进一步增强模型的泛化能力。另外,建立了愉悦、困惑、惊讶、中性和疲倦5种情感状态的模型,并依据所提出的模型进行了实际验证实验,实验结果表明所提出的优化后的三维状态空间情感模型相比于传统模型识别准确率提升了12.5个百分点。The traditional emotion model only focuses on the relationship between student learning expressions and corresponding learning emotions,but ignores the relationship between different learning emotions,thus resulting in relatively low accuracy of student learning expression recognition.Based on this,we establish a 3D state space emotion model of student learning expressions and introduce Maxout neurons into it to build an optimized 3D state space emotion model,further solve the 3D gradient dispersion problem,and better optimize the training process of the system.The generalization ability of the model is further enhanced.In addition,five models of emotion states,such as pleasure,confusion,surprise,neutrality and fatigue,were established,and practical validation experiments were conducted based on the proposed model,and the experimental results showed that the optimized 3D state-space emotion model proposed in this paper improved the recognition accuracy by 12.5 percentage points compared with the traditional model.

关 键 词：学生学习情感模型 三维状态空间情感模型 Maxout神经元 情感分类器 学生学习表情识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]