Abstract：Geocentric motion sequence contains complex noise and the real signal is difficult to extract effectively, so we propose a noise reduction method LOESS-EMD combining the locally weighted regression(LOESS) and the empirical mode decomposition(EMD) methods. Firstly, we fit the geocentric motion sequence with the LOESS method to obtain the fitted time series and the residual sequence. Then, we perform the EMD method on the residual series and extract the low frequency signal from it. Finally, we reconstruct the fitted time series and the low frequency signal in the residuals to obtain the denoised time series. In the simulation data experiment, compared with the LOESS method, the LOESS-EMD method reduces the root mean square error by 31% and improves the signal-to-noise ratio and percentage of remaining energy by 16% and 0.16 percent point. This method is used to denoise the geocentric motion sequence provided by the International GNSS Service(IGS) 3rd reprocessing campaign(Repro3), and the experimental results show that the LOESS-EMD method can effectively reduce the noise of the geocentric motion sequence and thus improve its accuracy.
KE Neng,ZHU Xinhui,WANG Ren et al. A Noise Reduction Method for Geocentric Motion Based on Locally Weighted Regression and Empirical Modal Decomposition[J]. jgg, 2023, 43(9): 904-908.
