表情机器人的动态情绪调节过程研究

Dynamic regulation process of facial expression robot

本文提出了一种基于概率有限状态机的表情机器人情绪表现模型,将其应用到实时动态调节的表情机器人面部表情上.为实现该模型,首先定义表情机器人的情绪状态空间,并通过调查获取不同情绪状态的刺激转移概率.结合Gross的情绪调节过程,抽象出情绪表现规则中的抑制特征因子和人机交互关系因子,并使用遗传算法对其进行优化,同时采用自适应变异概率算子和交叉算子对优化过程进行实时的调节,其参数性能得到了相应的提高对模型参数进行了量化研究及交互效果的仿真分析,并在所研制的23自由度表情机器人平台上进行了相关实验此外,对于实际交互效果,还进行了统计学的调查分析.结果表明,本模型能够摆脱单一的表情交流方式,得到符合当前交互环境的表情.

This paper deals with the emotion state-space model and the implementation of robot facial expression based on the probabilistic finite-state machine for the dynamic emotion regulation. The emotion state space is defined and the stimulating transition probabilities of different emotion state are acquired. The inhibitory characteristic coefficient and the human machine relationship coefficient are merged with Grossian emotional regulation process. The corresponding performances of parameters are enhanced using the genetic algorithm optimization and the real-time regulation of self- adaptive mutation probabilistic operator and cross-over operator. The quantitative analysis of the model parameters is made. The results generated by the emotion expression model are verified using the 23-degree-of-freedom expression robot platform. Moreover, the interactive effects are analyzed by the statistical algorithm. It also shows that the emotion expression model can acquire online expressive results and get rid of the single expressive interaction mode comparing to traditional algorithms.