A Real Signal Model-Based Method for Processing Boundary Effect in Empirical Mode Decomposition

The boundary effect is an important issue in empirical mode decomposition (EMD). It makes the decomposition results inaccurately and inefficiently. Since the boundary points cannot be determined, boundary effect exists in the procedure of decomposition of EMD. For resolving this issue, a real signal model-based method is proposed in this paper. Based on the real signal model, the boundary points could be treated as 0 points when interpolating upper and lower envelops. For making a signal conforms to this model, a window function which has 0 values at boundary points is needed. In order to ensure the window function play an AM role for each frequency component, the frequency of the window function must be very small. After performing EMD with the windowed signal, the intrinsic mode functions (IMFs) of the original signal can be reconstructed accurately. Simulation experimental results show that the proposed method is better than mirror symmetry, mirror symmetry after extension and cosine window-based method. Although the proposed method may produce fake frequency components with the very low frequency component, the comparison results still show that our method is much better than other methods.