Evaluation of EMD-based Methods on GPR Data

Empirical Mode Decomposition (EMD) is a relatively new technique for analysing non-linear and non-stationary time series. Ensemble Empirical Mode Decomposition (EEMD) and Complete Ensemble Empirical Mode Decomposition (CEEMD) are noise assisted and adaptive methods based on EMD. Here, we compare the empirical mode decomposition methods using both synthetic and real GPR data. In particular we examine: (1) the separation of high frequency wavelets from the low frequency ones and (2) the noise level that yields better decomposition for EEMD and CEEMD. We also examine the capability of these decomposition methods to remove random and coherent noise on real GPR data.