Abstract:We propose a new EMD denoising method to solve the problem that the K value of the demarcated intrinsic mode function (IMF) cannot be determined directly in the process of Empirical mode decomposition (EMD) denoising when the number of IMF components of high frequency noise is less than the number of low frequency IMF components. This method uses an average period and energy density product index method to automatically determine the K value of the demarcated IMF, reconstructs the IMF components of high frequency noise, and subtracts the reconstructed noise from the original signal. The method is verified using the simulated data and the measured GPS elevation time series data of BJFS station. The experimental results show that the proposed method can directly determine the K value of the demarcated IMF and reduce the computational complexity. It was more reliable than the traditional EMD method in the noise reduction of GPS elevation time series.