一种基于不确定方法的虚拟人眨眼控制模型

Uncertainty method for control of virtual human's blink

针对采用确定性方法合成的虚拟人眨眼效果单一、重复、不能刻画眨眼表达深层信息的不确定性,提出一种基于正态云模型的虚拟人眨眼非确定性控制方法。设计了眨眼间隔云、眼皮旋转速度云、睁眼最大角度云和闭眼保持时间云。分别运用正向云生成算法产生眨眼间隔、眼皮旋转速度、睁眼最大角度和闭眼保持时间,且四个参数符合各自云模型的模糊性和随机性规律。算法分析和与确定性方法的对比表明,所提方法能够生成具有随机差异的虚拟人眨眼控制律。给出了四类眨眼云模型数字特征的建议值。分析了通过显著改变眨眼云数字特征值实现常闭眼偶尔睁眼的不确定性控制的可行性。

Blink synthesized base on certainty methods for virtual human is simple and repetitive.They cannot Describe uncertainties of blink expressing deep mind.This paper proposed a normal cloud model based uncertainty method for control of virtual human＇s blink.Designed four cloud models,which were blink interval cloud,eyelid rotation cloud,maximal eye＇s open angle cloud,and eye＇s keeping close time cloud.With direction of the four clouds,blink interval,eyelid rotation speed,maximal eye＇s open angle,and eye＇s keeping close time were generated.The four parameters follow fuzziness and randomness of respective clouds.Results of algorithm analysis and comparison with certainty methods show that the proposed cloud model based uncertainty method could generate control law with randomness for blink actions of virtual human.Proposed feature values of the four kinds of blink clouds.Analyzed the feasibility to realize uncertainty control of eyes＇ close state companied with occasional open,by obviously changing feature values of the four blink clouds.