Inhibition Of The Empirical Mode Decomposition End Effect Of Commonly Used Methods Of Performance Comparison Study

In 1998, Huang et al. proposed empirical mode decomposition (EMD) to estimate the Hilbert spectrum of a signal. This new signal processing method based on the EMD and Hilbert spectrum can get the instantaneous information from nonlinear and non-stationary signals, which makes the signal analysis more accurate. In essence, this method processes a signal, and makes it become stationary, which results in the decomposition with different scales or fluctuations in the trend, produce a series of modes called intrinsic mode function (IMF) . This can make the Hilbert transfer of IMF'instantaneous frequency get the true physical meaning.However, this method also has limitation when the upper and lower envelopes, which are constructed by using a cubic spline function from given signals, deviate from the true ones because of the unknown accurate data at the end of the signal. This defect influences the signal decomposition quality with EMD algorithm seriously. In order to solve this problem, several approaches for restraining the end effect were proposed, and all have a certain degree of restraint effect. However, it is a lack of a systemic and accurate evaluation for the performance of these approaches. In this paper, several commonly used restraining methods, which include Matching Extending, Mirror Extending, Symmetrical Extending, and Polynomial Fitting, have been evaluated by comparing. It used EMD which is treated by these methods to decompose non-periodic and non-stationary signal. Through analyzes the respective intrinsic mode function, estimates the respective Hilbert spectrum, draws up the contrast chart of the respective superposed signal of main intrinsic mode function and the primary signal, and the contrast chart of the average frequency curve and the standard frequency, calculates the mean error of the respective superposed signal of main intrinsic mode function and the primary signal, and the mean error of the average frequency curve and the standard frequency, as well as computes complexity and draws up the time contrast histogram of each method and so on, this paper comparely studied each method performance carefully. Finally summarized the serviceability of each method systematically, the suppressing effect of each method, as well as the good and bad points of each method.