三维人体骨骼动画自动合成方法研究

Research on Automatic Synthesis Method of Three-Dimensional Human Skeleton Animation

传统的人体骨骼动画制作方法是参考真实人体骨骼运动过程中各关节点的坐标位置变化轨迹,确定关键帧与插值帧模型,再进行必要的编辑处理,这需要大量的专业领域知识以及复杂的交互规则,人力物力成本高。为解决上述问题,提出了一种在给定任意两个人体骨骼动作数据作为首尾帧的条件下,自动合成完整三维人体骨骼动画序列的方法。方法由基于卷积自编码网络的骨骼动画特征提取模型和双线性插值与卷积相结合的插值帧自动合成模型两部分组成。随机从Mocap数据库中抽取两帧人体动作数据作为模型的输入,可以自动生成三维人体骨骼动画。文中方法与传统插值帧生成方法相比,具有较好的动作趋势转折信息的预测和还原能力,提升了动画系统的交互效率以及智能水平。

The traditional method of human skeleton animation is to refer to the coordinate position of each node in the real human skeleton movement, determine the keyframe and interpolation frame model, and then make the necessary editing. This requires a lot of specialized field knowledge and complex interaction rules and often requires a lot of manpower and material costs. To solve this problem, a method of automatically synthesizing missing frame data between the first and last frames is proposed in this paper. The method consists of two parts: the feature extraction model of skeletal animation based on convolution self-coding and the automatic synthesis model of interpolation frame combined with bilinear interpolation and convolution. Random extraction of two frames of human movement data from the Mocap database as input to the model can automatically generate three-dimensional human skeleton animation. Compared with the traditional interpolation frame generation method, this method has a good ability to predict and restore the movement trend transition information. It improves the interactive efficiency and intelligence level of the animation system.