情感虚拟人研究
发布时间：2018-11-01 21:27
【摘要】： 情绪在人类的日常行为和感官体验中处于核心地位,人类与当前智能机器的最大区别就在于人类具有产生情绪的能力。不断发展的情绪理论有力地支持了计算机科学中人工智能领域的研究。越来越多的人工智能学者认识到,为了实现计算机“拟人化”的美好梦想,仅仅使其具有智能是远远不够的,还必须让计算机理解并具有类似人类的情绪,能与人进行情感的交互。 本文主要研究人工情感的理论和情感计算的实现方法,设计了一个具有个性的虚拟人情感模型,并开发了相应的系统。该系统主要包括:(1)文本处理与响应产生模块:采用人工智能与自然语言理解的相关技术提取用户输入的文本中的情感信息;(2)性格模型:采用人格特质的五因子模型定义虚拟人的性格空间,依据设定的性格类型,结合当前的情感状态,采用贝叶斯推理规则得到下一时刻的情感状态;(3)情感—情绪模块:结合由前两个模块获得的情感信息,定义了情绪转移矩阵,计算每一种情绪的转移概率,并把其中概率的最大者作为下一时刻的情绪状态;(4)同步模块:主要给出从情绪状态到表情的映射。当两种情绪的概率都大于阈值时,同步模块决定对这种情况的取舍,并决定在过渡情绪之间出现什么样的表情。 最后编程实现了这个情感虚拟人交互系统。对系统测试结果的分析表明,由这个模型模拟的情感反应是符合人类情感行为规律的,验证了本文所提出的情感计算的模型。
[Abstract]:Emotion is at the core of human daily behavior and sensory experience. The greatest difference between human beings and current intelligent machines lies in the ability of human beings to produce emotions. The evolving emotional theory supports research in the field of artificial intelligence in computer science. More and more artificial intelligence scholars realize that in order to realize the beautiful dream of "humanization" of computers, it is not enough to make them intelligent, but also to make computers understand and have a similar human emotion. Be able to interact emotionally with people. In this paper, the theory of artificial emotion and the realization method of emotion calculation are studied, a virtual human emotion model with personality is designed, and the corresponding system is developed. The system mainly includes: (1) text processing and response generation module: using artificial intelligence and natural language understanding technology to extract emotional information from the text input by users; (2) Personality model: the five-factor model of personality trait is used to define the personality space of virtual human. According to the set personality type and the current emotional state, Bayesian reasoning rule is used to obtain the emotional state of the next moment; (3) emotion-emotion module: according to the emotional information obtained from the first two modules, the emotional transfer matrix is defined, the probability of each emotion is calculated, and the highest probability is regarded as the emotional state at the next moment; (4) synchronization module: the mapping from emotional state to expression is given. When the probability of both emotions is greater than the threshold, the synchronization module determines the choice of the situation and the expression between the transition emotions. Finally, the virtual human interaction system is realized by programming. The analysis of the system test results shows that the emotional response simulated by this model is in accordance with the human affective behavior law and verifies the emotional calculation model proposed in this paper.
【学位授予单位】：上海师范大学
【学位级别】：硕士
【学位授予年份】：2008
【分类号】：TP181
本文编号：2305143
[Abstract]:Emotion is at the core of human daily behavior and sensory experience. The greatest difference between human beings and current intelligent machines lies in the ability of human beings to produce emotions. The evolving emotional theory supports research in the field of artificial intelligence in computer science. More and more artificial intelligence scholars realize that in order to realize the beautiful dream of "humanization" of computers, it is not enough to make them intelligent, but also to make computers understand and have a similar human emotion. Be able to interact emotionally with people. In this paper, the theory of artificial emotion and the realization method of emotion calculation are studied, a virtual human emotion model with personality is designed, and the corresponding system is developed. The system mainly includes: (1) text processing and response generation module: using artificial intelligence and natural language understanding technology to extract emotional information from the text input by users; (2) Personality model: the five-factor model of personality trait is used to define the personality space of virtual human. According to the set personality type and the current emotional state, Bayesian reasoning rule is used to obtain the emotional state of the next moment; (3) emotion-emotion module: according to the emotional information obtained from the first two modules, the emotional transfer matrix is defined, the probability of each emotion is calculated, and the highest probability is regarded as the emotional state at the next moment; (4) synchronization module: the mapping from emotional state to expression is given. When the probability of both emotions is greater than the threshold, the synchronization module determines the choice of the situation and the expression between the transition emotions. Finally, the virtual human interaction system is realized by programming. The analysis of the system test results shows that the emotional response simulated by this model is in accordance with the human affective behavior law and verifies the emotional calculation model proposed in this paper.
【学位授予单位】：上海师范大学
【学位级别】：硕士
【学位授予年份】：2008
【分类号】：TP181
【引证文献】
相关期刊论文 前1条
1 赵涓涓;杨建峰;陈俊杰;王玉友;;基于情绪认知评价理论和Q-learning的人机交互中情感决策[J];太原理工大学学报;2012年06期
相关硕士学位论文 前2条
1 王若琳;电视人物专访节目中的心理互动[D];吉林大学;2011年
2 王玉友;基于情绪认知评价理论的人机交互中情感交互研究[D];太原理工大学;2012年
，本文编号：2305143
本文链接：https://www.wllwen.com/kejilunwen/rengongzhinen/2305143.html
