一种自适应二阶指数平滑滤波在Kinect骨骼提取的应用

Application of an Adaptive Two Order Exponential Smoothing Filter to Kinect Skeleton Extraction

为了改进Kinect获取骨骼坐标的稳定性,提出了一种基于二阶指数平滑滤波的方法,引入移动速度趋势增量,并根据当前速度情况,采用一次线性插值方法实现平滑系数的自适应化,最后通过关节角度计算来获取人体上半身的关键部位角度,并映射到仿人机器人本体。实验结果表明,该算法拥有更好的平滑过渡性,并且能够有效地避免由于原始数据突变带来的仿人机器人手臂抖动的情况。

In order to improve the stability of the skeleton coordinates obtained by Kinect, this paper puts forward a method based on second order exponential smoothing filtering. In this paper, the moving speed trend increment is introduced, and a linear interpolation method is adopted to realize the adaptive smoothing coefficient according to the current speed. Experimental results show that the proposed algorithm has a better smooth transition, and can effectively avoid the humanoid robot arm jitter caused by the mutation of original data.