低采样率下经验模态分解性能提升研究  

Performance improvement of EMD under low sampling rates

作　　者：黎恒[1] 李智[2,3] 莫玮[1] 

机构地区：[1]西安电子科技大学机电工程学院,西安710071 [2]桂林航天工业学院自动化系,桂林541004 [3]桂林电子科技大学电子工程与自动化学院,桂林541004

出　　处：《振动与冲击》2016年第17期185-190,共6页Journal of Vibration and Shock

基　　金：国家自然科学基金(61361006)

摘　　要：经验模态分解(EMD)使用信号极值点的位置和取值信息进行分解,对采样率有较高的要求。针对EMD在低采样率下性能降低的现象,提出一种基于B样条拟合的信号局部均值计算方法。首先提取信号极值点出现的时刻作为尺度,然后通过对极值点时刻进行重新采样构造B样条节点,最后应用B样条最小二乘拟合方法直接计算局部均值。与EMD方法相比,该方法不需要信号极值点的准确位置和取值,因此不容易受到低采样率的影响。对平稳信号和非平稳信号的仿真结果表明,该方法能在接近奈奎斯特频率的低采样率下获得较高的性能。与基于插值的解决方案相比,该方法的分离性能更好。Empirical mode decomposition（ EMD） depends highly on exact locations and values of signal extrema,they need a higher sampling rate. Aiming at improving the performance of EMD under low sampling rates,a local mean estimation method based on B-spline fitting was proposed. Firstly,the occuring instant of extrema was extracted as a time scale. Then,the occuring instant of extrema was re-sampled to construct node points of B-spline. Finally,the local mean was computed directly with B-spline least squares method. Compared with EMD method,the proposed method did not need the exact locations and values of extrema,so the low sampling rates were not easy to affect this technique. The efficiency of this technique was demonstrated using simulated signals. The simulation results showed that the performance of the proposed method is not reduced even the sampling rate is close to Nyquist rate; the proposed method is superior to existing interpolation methods in separation performance.

关 键 词：经验模态分解 B样条拟合 低采样率 信号分解 时频分析 

分 类 号：TH137[机械工程—机械制造及自动化]