Computing with Words for Understanding and Generation of Modal Human Motions affected by Mental and Physical Conditions

在日常的人类运动中，他们可以表达主要的目标运动，并在任何时刻受到他们的心理和/或身体条件的影响进行二次修改运动。为了实现人性化的交互系统，引入人体运动理解和/或生成的人机界面是至关重要的。本研究提出了一个概念框架，人机界面使用人体运动的人-智能机系统，1）“理解人体运动的仿人认知机制”它不仅可以从人体运动中区分出原始运动，而且可以从二次修正运动中理解情感的心理和生理状态。2）“虚拟人体的运动修正机制”它使智能机器不仅能产生虚拟人体的初级目标运动，而且能产生嵌入一定心理和生理条件的虚拟人体的次级运动， 关于我们 这些机制的方法采用概念计算与文字捕捉人类的行动目标和不确定的心理和身体条件，作为人类可理解的语义复合体。为了更具体地实现物理运动特征、模态运动的语言类别以及心理和物理条件的语言类别之间的相互转换，该方法采用了一些特定的软计算过程，即，参数特征空间方法，隐马尔可夫模型，Choquet积分代理网络和GA。为了验证该方法的有效性，进行了一些实验，其中受试者播放几个给定的日常运动与几个给定的模态修改。通过三维运动捕捉系统测量的实验数据被应用到所提出的机制与文字计算。然后，估计的运动模态被证实是相当有效的关于人类的模仿认知机制，理解人类的运动。对于给定的几个主运动和几个给定的模态修改，通过运动修改机构，验证了虚拟人台运动的生成是足够的。少

In everyday human motions they can express primary objective motions with secondary modification motions affected by their mental and/or physical conditions at any moment. For realizing human friendly interactive systems, introduction of human interfaces using body motion understanding and/or generation is crucial.This study proposes a conceptual frameworks on human interfaces using human motions for man-intellectual machine systems, and shows the ways to realize the mechanisms mentioned below.1) "Human mimetic cognitive mechanisms for understanding human motions" From human motions it can not only distinct the primary motions but also understand the affective mental and physical conditions from the secondly modification motions.2) "Motion modification mechanisms of virtual mannequins" It makes intellectual machines generate not only the primary objective motions but also the secondly motions of the virtual mannequins which are embedded certain mental and physical conditions.The presen … More t approach to these mechanisms employs the concept computing with words to capture human action objectives and uncertain mental and physical conditions as human comprehensible semantic complexes. To realize more specifically mutual transformations among physical motion features, linguistic categories for modal motions, and linguistic categories for mental and physical conditions in this approach, it employs some specific soft computing procedures, i.e., parametric eigen-space methods, hidden Markov models, Choquet integral agent networks, and GA.To verify the effectiveness of the approach some experiments were conducted, where the subjects played the several given everyday motions with several given modal modifications. The experimental data measured by a 3D motion capture system were applied to the proposed mechanisms computing with words. Then the estimated motion modalities were confirmed to be rather valid concerning the human mimetic cognitive mechanisms for understanding human motions. And the generated motions of virtual mannequin were confirmed to be adequate concerning the motion modification mechanisms for several given primary motions with several given modal modifications. Less