【作者】 金钺；

【导师】 姚福生； 侯文君；

【作者基本信息】 北京邮电大学 ， 机械电子工程， 2008， 博士

【摘要】 情感计算是人工智能和认知科学一直关注的难题,也是创建和谐人机交互环境的重要内容。如何构筑一个能进行情感识别和生成的类生命体是情感计算和人工智能的热点问题。由于目前的虚拟学习环境或远程教育系统缺乏表现学习属性的个性化情感的交互,实际效果与传统面对面教育有明显的差异,加之情感计算学科也缺乏对于情感产生完整过程的研究与表示,因此本文以虚拟学习环境为情感计算的研究载体,以人工心理和情感计算理论为导向,通过心理分析、行为调查、功能映射、模拟计算等手段,进一步探讨了情感产生与发展机制、学习动机及其与学习情感的映射关系、情感模拟过程以及智能Agent实现方法等问题,为提高虚拟学习环境的教学效果及促进教育信息化发展提供相关理论与技术支持。具体研究内容如下:(一)提出了计算机处理情感的基本流程(FAC——flow ofaffective computing)。首先提出将计算机模拟情感视为一个完整的事件过程的思想,然后对情感的产生、发展、表现的一系列具有本质联系的连续流程以及其驱动、运行原因进行深入的探讨与研究,最后综合情绪心理学领域的相关研究成果,针对情感计算的特点与需求,分析、概括出情感产生发展的机理,为情感建模提供有利的理论支持。(二)建立了情感与学习的映射关系。论述性格属性就是动机系统——即情感的个性化模型。在学习过程中这种动机系统称为学习动机。学习动机反映个体的性格属性,决定了学习情感的个性化。将学习情感视为学习动机基础与人格系统,并将其作为情感个性化的底层,通过论证两种典型学习动机理论的意义以及在虚拟学习环境中的实用价值,并根据其含义与特点建立学习动机模型——以成就动机理论作为模型的整体框架,以驱力论模型作为细节算法,用于具体计算不同学习者的学习动机。(三)根据FAC理论建立具体的基于认知规则的FAC模型结构,将处理节点按照功能特性封装在FAC模型各模块,并对模块进行详细的设计与框架搭建,用于模拟人的学习过程中的情感激发过程。每个模块用于表示情感激发过程的各个节点,它根据上层输入与学习者个性及情感状态得到不同的处理结果。考虑学习者的个体属性,通过大量实验与测试,经过BP方法拟合得到个体的学习动机,并以此构建人格系统,从而使FAC模型能够激发个性化情感。(四)以几个最具代表性且受到广泛研究的Agent-BDI设计思想为基础,提出适应学习过程的基于FAC模型的Agent实现方法,该方法是一套适合FAC模型的、易于实现的、符合情感激发本质的Agent实现方法,用于实现各Agent功能模块,解决多Agent间的响应与交互问题,并将多Agent组织结构进行规划与设计,依据此方法对FAC模型中的个体Agent进行建模与实现。最后,结合研究成果与理论特点,选取了相关算法与技术手段,进行了程序实现与验证,将FAC-Agent最终构筑成一个能进行自主情感识别和生成的类生命体——虚拟人,它可以随学习者在虚拟环境中一起进行学习,并模拟出学习者真实的情感变化。同时对论文所作的研究工作进行了总结,指出了本论文所需进一步研究的问题和内容。更多还原

【Abstract】 Affection is essential to the learning process; Positive feelings can greatly improve the efficiency of learners and learning effectiveness. Excellent teachers can learn students’ state of study and grasp of knowledge timely by aware of the feelings. But also through the affective exchanges of students to encourage or supervision, so as to enhance good feelings, negative sentiment weakened. Due to the lack of affective interaction, there is a significant difference between traditional face-to-face education and the practical effect of the virtual study environment or long-distance education system. How can teachers access the learners the students’ affections, making use of affective remote teaching in virtual study environment, are currently the issues requiring urgent solution.This article based on the feelings in the learning process and the important role of the virtual learning environment and significance of the realization focuses on the learner’s achievement motivation, affection in computer simulation methods and structures construction of the Agent based on learning and affective problems-flow of affective computing (FAC). In-depth study of the affections generated and development mechanism, motivation to learn and its affective mapping relationship, achieving simulation framework of affection, Agent methodology based on learning and affective, affective computing theory and modeling and learning motivation model based on the affective psychology and their implement model are advanced to the and affection in a virtual learning environment in the implementation model. On the basis of this researches appealing Agent methodology to adapt to study process. FAC model of individual is implemented based on this methodology. Agent technologies based on affective computing in virtual learning environment, is a bold innovation, and plays a role as possible. This will greatly improve the virtual learning environment and the practicality of distance education system. The specific contents are as follows:(A) Computer simulation for affection will be made as an integrity of the process, as well as affective creation, development, performance of a series of continuous nature of the processes associated with driving, running reasons for the in-depth study and research, comprehensive The Psychology of Affection - related research results, affective computing against the characteristics and needs, analysis, have summarized the affective development of the mechanism and advanced flow of affective computing (FAC) to explore new computers to handle the basic affective process, affective modeling for providing the theoretical support.(B) This paper demonstrates the significance of two typical study motivation theories and practicality in the virtual learning environment, constructs a model of study motivation in accordance with its meaning and characteristics - achievement motivation theory as the overall framework of model and drive model as the detailed algorithm. According to the theory of affective computing flow based on psychology of affection, the affective is viewed as motivation which composes a learner’s personality, mapping with learning model, used in a virtual learning environment for learners in the affection generated simulation and recognition.(C) The use of existing multi-agent technology and learning theory will achieve the model of affective computing in the computer, single-agent structure in the system of individual learners used to simulate learner in the learning process affective excitation process, which is same as the process of the FAC. In this final FAC-agent system will be constructed to form affection-driven virtual learning environment through detailed design and implementation for agent’s system and function modules. Considering the individual properties, through experiments and tests, after BP fitting way by the individual motivation to learn and to build character, so that FAC model can stimulate individual feelings.(D) Making use of several most representative and extensive Agent-BDI-based methodology, Agent FAC methodology is advanced to adapt to the learning process, the methodology is an appropriate learning environment, easy to achieve, in line with affection excited nature of the Agent methodology used to achieve the Agent module, solve the communications problem of response and interaction amang multi-Agent, design multi-Agent organizational structure for planning and design. Individual FAC-Agent can be implemented based on the methodology.Finally, the combination of theoretical research results and characteristics, select the relevant algorithms and technical means to carry out the program and verification so that the FAC-Agent can be as learners in a virtual environment for learning and simulate learners’ the real affective change. At the same time, the paper conducted a study concluding that the present papers required further study the issue and content.更多还原