基于正态云模型的虚拟人点头动作不确定性控制

Uncertainty control based on normal cloud model for nodding actions of virtual human

设计了点头次数肯定云、最大点头角肯定云和点头速度肯定云。运用各自的正向云生成算法指导产生一个点头动作序列的点头次数、最大点头角度和点头速度。三个参数以各自云模型的期望值为中心正态分布。提出的CBNCA算法能够产生组间呈规律性差异的点头动作序列控制曲线。为描述点头动作序列组内差异性,提出了CBNCA+算法,同时实现组间和组内点头动作的不确定性控制。对比实验验证了两个算法控制点头动作不确定性的有效性,并给出了三类云模型数字特征的建议值。

Three kinds of cloud models were designed, which are the nodding times approving cloud, the maximal nodding angle approving cloud and the nodding speed approving cloud. With the guidance of respective positive cloud generation algorithms of the three models, three parameters in a nodding action serial, which are the nodding times, the maximal nodding angle and the nodding speed, were generated. The three parameters follow normal distributions around expected values of the three cloud models respectively. The proposed CBNCA algorithm generates control curves of nodding actions with disciplinary differences between different nodding action serials. In order to describe differences of actions in one nodding action serial, CBNCA＋ algorithm was proposed. It can realize uncertainty control of nodding action both between different action serials, and between different actions in one serial. Experiment results prove the effectiveness of the uncertainty control for nodding action of the two cloud-based algorithms. Feature values of the three kind cloud models were proposed.