A Method On Preferential Dominant Frequency And Mount Edge Filter Decomposing Signal Into Intrinsic Mode Functions

EMD is a very important part in empirical mode decomposition and Hilbertspectral analysis (EMD/HSA). Fast filtering decomposes signal into IMFs (FFDSI)method based on the discrete Fourier transform, and decomposes signal into IMFs byrectangular filter in frequency domain. It is fast in computing and easy to use, besides,the IMFs and the residual gotten by FFDSI are strictly orthogonal each other. Thus,FFDSI can ensure the signal has the same energy before and after decomposition.However, FFDSI has some shortcomings as the following:(1) As FFDSI decomposing signal into IMFs by rectangular filter, there must bea serious effect of Gibbs resulted that can make the IMFs anamorphic.(2) Rectangular filter will weaken the transient characteristic of the IMFs.(3) The way that FFDSI filtering according to sequence of high frequency to lowis very possible to separate the "beats" the signal has and make IMFs’ physicalmeaning undefined.To response to these issues, in this paper a method on preferential dominantfrequency and mount edge filter decomposing signal into IMFs is given. The newmethod replace rectangular filter with mount edge filter in FFDSI, and determine thefiltering order according to the absolute value of discrete Fourier transform to get theIMFs.The method on preferential dominant frequency and mount edge filterdecomposing signal into IMFs has the following advantages:(1) The method by using the mount edge filter can effectively reduce the Gibbseffect generated by a rectangular filter, so the decomposition results is more objectiveto show the true frequency components contained in the original signal.(2) The method on preferential dominant frequency and mount edge filterdecomposing signal into IMFs can reflect the transient nature of the original signaland the characteristics of time-frequency analysis of EMD/HSA better.(3) The new method filtering will no longer follows from high-frequency tolow-frequency, instead, the principle of preferential dominant frequency be used. Itcan effectively prevent the "beats" from be separated. So the physical meaning of thedecomposition results is easier to understand.From the above, the new method is just to make up for the shortcomings ofFFDSI method.