【作者】 李雪；

【导师】 赵姝颖；

【作者基本信息】 东北大学 ， 电子信息， 2023， 硕士

【摘要】 科技创新和科学普及是实现创新发展的两翼,科学普及工作是我国实现高水平科技自立自强的一项重要基础性工作。而AI普及教育是科学普及工作中的关键任务,因此,本文将所研究的基于情感识别的人机交互技术应用于AI普及教育中,开发了用于AI普及教育的情感识别互动教学系统,并在全国科普教育基地中进行科学普及相关工作。在实现科技创新的同时达到科学普及的目的。本文主要工作如下:(1)构建了基于面部图像的情感识别模型,并针对数据集的不均衡性导致的模型对不同类别的识别准确率差异较大的问题,进行了数据均衡化方法研究。研究方法包括在原数据集上使用过采样方法和欠采样方法降低数据的不均衡程度,以及使用权重平衡法调整模型对少数类和多数类的关注度。实验结果表明,本文所提的过采样方法能够有效提升少数类准确率及模型整体性能。(2)开发了基于情感识别的人机交互应用。首先构建了用于交互的情感语义文本库,然后设计了交互应用的功能,并将面部情感识别模型转换为可执行程序。最后基于“汉语言”平台设计了应用的交互界面,实现了名为“情感对对碰”的情感交互应用。(3)设计并实现了用于AI普及教育的情感识别互动教学系统。系统由知识教学模块、情感交互学习模块和知识测评模块组成。在知识教学模块中,学生可以学习AI知识,并通过模块中的课堂交互内容加深对知识的理解。情感交互学习模块即第二部分的情感交互应用,通过该模块,学生可以随时随地体验人工智能技术,将学习到的知识具象化。知识测评模块可以对所学内容进行测试,通过情感识别的方式进行答题,既巩固了学习内容,又增强了学习趣味性。该系统既可以作为AI普及教育的课堂教学应用,也可以在科普教育基地中进行展出,将本文的科研成果转化为科普教育资源,达到科学普及的目的。更多还原

【Abstract】 Scientific and technological innovation and the popularization of science are two wings of realizing innovative development.The popularization of science is an important foundational work for realizing self-reliant self-improvement in high-level science and technology.The popularization of AI education is a key task in the popularization of science.Therefore,thesis applies the human-computer interaction technology based on emotion recognition to the popularization of AI education,develops the interactive teaching system of emotion recognition for the popularization of AI education,and carries out the related work of science popularization in the national science popularization education base.In the realization of scientific and technological innovation at the same time to achieve the purpose of science popularization.The main work of thesis is as follows:(1)The emotion recognition model based on facial images is constructed,and the data equalization method is studied to solve the problem that the recognition accuracy of different categories of models is greatly different due to the unbalance of data sets.The research methods include oversampling and undersampling methods on the original data set to reduce the degree of unevenness of the data,and the use of rebalancing methods to adjust the model’s attention to minority and majority classes.Experimental results show that the oversampling method proposed in thesis can effectively improve the accuracy of a few classes and the overall performance of the model.(2)Developed human-computer interaction application based on emotion recognition.Firstly,the emotional semantic text base for interaction is constructed,then the function of the interactive application is designed,and the facial emotion recognition model is transformed into an executable program.Finally,the interactive interface of the application is designed based on the"Chinese language" platform,and the emotional interaction application named"emotional pair up" is realized.(3)An emotion recognition interactive teaching system for AI popularization education is designed and implemented.The system consists of a knowledge teaching module,an emotional interactive learning module,and a knowledge assessment module.In the knowledge teaching module,students can learn AI knowledge and deepen their understanding through the classroom interactive content in the module.The emotional interaction learning module is the second part of the emotional interaction application,through which students can experience AI technology anytime and anywhere and visualize the knowledge they have learned.The knowledge assessment module can test the learned content and answer questions by means of emotion recognition,which not only consolidates the learning content but also enhances learning fun.The system can be used as a classroom teaching application for AI popularization education,and can also be exhibited in science education bases,transforming the scientific research results of thesis into science education resources and achieving the purpose of science popularization.更多还原