基于CEEMDAN和小波包多阈值的GNSS高程时间序列去噪方法

为提高GNSS高程时间序列的去噪效果，以仿真信号和拉萨站2000~2020年高程时间序列为例，采用自适应噪声完备集合经验模态分解（complete ensemble empirical mode decomposition with adaptive noise，CEEMDAN）方法将信号分解成若干个特征模态函数（intrinsic mode function，IMF），对每个IMF分量进行小波包多阈值分解，依据不同节点能量占IMF总能量百分比选择不同的阈值准则，将降噪后的节点重构得到降噪后的IMF分量，进而得到降噪后的时间序列。利用信噪比、均方根误差等指标对比分析本文方法、EMD、CEEMDAN、小波去噪和小波包多阈值去噪等5种方法的去噪效果。结果表明，本文方法效果最优。

GNSS Height Time Series Denoising Method Based on CEEMDAN and Wavelet Packet Multi-Threshold

To improve the denoising effect of GNSS height time series, we decompose the simulation signal and height time series of Lhasa station from 2000 to 2020 into several intrinsic mode functions(IMF) by complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN) method respectively. We perform wavelet packet multi-threshold decomposition for each IMF component, select different threshold criteria according to the percentage of the energy of different nodes in the total energy of the IMF, reconstruct the noise-reduced node to obtain the noise-reduced IMF component, and then obtain the noise-reduced time sequence. By the indexes of signal-to-noise ratio and root mean square error, we compare and analyze the denoising effect of the proposed method, EMD, CEEMDAN, wavelet denoising and wavelet packet multi-threshold denoising. The results show that the proposed method has the best effect.