虚拟数字人手势交互设计

Gesture Interaction Design of Virtual Human

目的设计一个通用、动态、自学习的虚拟数字人手势生成模型,为生成丰富情感表现力的虚拟数字人提供数据和模型基础。方法首先总结隐喻手势的语义分类方法,构建手势的量化描述语言,为手势的语义可计算提供量化方法;然后构建针对不同应用场景的情感语料库,标注手势的情感、手势隐喻语义和手势的量化描述,挖掘不同场景下虚拟数字人的行为模式,为虚拟数字人手势生成研究提供数据基础;最终提出基于应用场景的虚拟数字人手势生成算法,建立抽象的交流意图与手势的物理实现之间的映射,为生成具有环境认知能力的虚拟数字人提供模型支持。结果通过两个案例验证了该方法的有效性,解决了虚拟数字人对环境认知和互动不足的问题,使虚拟数字人具有了行为的交互性和思想的智能性。结论研究成果可推动虚拟数字人技术的发展,为虚拟数字人在不同场景下的广泛应用提供数据和模型基础,具有广泛的学术意义和应用前景。

In order to design a general, dynamic, self-learning virtual human gesture generation model and provide data and model basis for the generation of virtual human with rich emotional expression, this paper first summarizes the semantic classification methods of metaphorical gestures, constructs a quantitative description language of gestures, and provides a quantitative method for the semantic computability of gestures. Then, the emotional corpus is built for different applications. It marks the emotion of gestures, the metaphorical semantics and quantitative description of gestures, and mines the behavior patterns of virtual human in different scenarios so as to provide a data basis for the virtual human gesture generation. Finally, a virtual human gesture generation algorithm based on application scenarios is proposed to establish the mapping between abstract communication intention and physical implementation of gestures, which provides model support for the generation of virtual human with environmental cognitive ability. As a result, the effectiveness of the method is verified by two cases, which solves the problem of virtual human’s insufficient cognition and interaction to the environment. The method makes virtual humans have the interaction of behavior and the intelligence of thought. This research achievement can promote the development of virtual human technology, provide data and model basis for the wide application of virtual human in different scenarios, and has a wide range of academic significance and application prospects.