虚拟人手语动画自动生成技术研究

Research on Automatic Generation Technology of Virtual Human Sign Language Animation

虚拟人手语动画中信息量较多,其手势动作复杂性较高。在手语动画生成过程中,通常缺乏动作约束能力,动画生成后具有重合部分且手部画面异常,导致手语动画生成精度低、质量差。为此,研究了虚拟人手语动画自动生成技术。根据动作数据采集结果,构建无向加权图,确定各个顶点位置与特征,得到原始动作捕捉序列,完成虚拟人手语动作分割,确定虚拟人手部动作基础信息。使用类函数与高度函数以及人体体征数据,提取虚拟人手部动作关键点,对手部特征点进行划分,分析手部动作的骨骼关节特征。使用线性变换公式以及动作处理帧窗口,对手语动画进行约束,实现手语动画自动生成。实验结果表明,本文方法可提升手语动画生成精度,实现手部滑动动作控制,提升虚拟人手语动画质量。

The virtual human sign language animation contains more information,and its gesture action is more complex.In the process of generating sign language animation,there is a lack of action constraint ability.After the animation is being generated,there are overlapping parts and abnormal hand images,resulting in low accuracy and poor quality of sign language animation.Therefore,the automatic generation technology of virtual human sign language animation is studied.According to the result of motion data acquisition,an undirected weighted graph is constructed to determine the positions and characteristics of each vertex,obtain the original motion capture sequence,complete the segmentation of the virtual human’s sign language motion,and determine the basic information of the virtual human’s hand motion.The class function,height function and human body sign data are used to extract the key points of the virtual human hand movement,divide the hand feature points,and analyze the skeleton and joint characteristics of the hand movement.Linear transformation formula and action processing frame window are used to constrain sign language animation and realize automatic generation of sign language animation.The experimental results show that the proposed method can improve the accuracy of sign language animation,realize the control of hand sliding motion,and improve the quality of virtual human sign language animation.