Abstract For the empirical mode decomposition (EMD) noise reduction process, there is signal and noise mode aliasing, and the intrinsic mode function (IMF) component is directly classified into high frequency noiseto generate real signals. The real signal is “submerged,” among other issues. This paper proposes an improved method of EMD noise reduction. The improved EMD denoising method mainly reconstructs the second IMF to demarcate the IMF component obtained by EMD decomposition, performs the next EMD decomposition on the reconstructed signal, obtains the “real” signal therein, and repeats this operation multiple times. Finally, all the obtained low frequency signals are accumulated to achieve the purpose of noise reduction. Experiments are carried out using three kinds of simulation data and one kind of GPS measured elevation data. The simulation data and the measured data are evaluated by different evaluation indexes. The results show that the improved EMD method has better noise reduction effects than the traditional EMD method, and the reliability of the method is verified.