Improved Algorithm And Application Investigation Of Empirical Mode Decomposition

Empirical mode decomposition (EMD, a new method of time-frequency analysisrecently introduced for processing non-stationary signals adaptively, demonstrates uniqueadvantages in non-stationary signal processingand possess important theoretical value andbroad application prospects. The purpose of this paper isto investigate the EMD algorithmfrom the perspective of system, and to discuss the relationship between cubic splineinterpolation and the number of extreme points, based on which a new method for restrictingend is proposed. Then the application of EMD to processing of helicopter acoustic signal andBlasting seismic signalis developed.From a system perspective, the EMD algorithm is seen as a multi-level gray-box systemseries, and cubic spline interpolation algorithm is a crucial step of EMD. Based on theanalysis of cubic spline interpolation, the standard deviationexpression of interpolationresults using different number of extreme pointsis given in the second class of boundaryconditions and not-a-knot boundary conditions. The relationship of extreme points tointerpolation results is discussed based on the standard deviationexpression.Carrying out two groups of simulation experiment using simulated signal and measuredsignal, the results verify the conclusions: when the increasing number of extreme points isgreater than a certain number, the interpolation results are not affected by increasing numberof extreme pointsBy comparison study of mirror symmetry method, data expansion method and datareconstruction method, their characteristics and efficiency aresummarized, and on this basisa improved EMD algorithm restricting endeffect is proposed. This method retainstheadaptation of data expansion method, while improving the operational efficiency. The EMDdecomposition results of simulated signals and measured signals shows that the method iseffective and high speedy.This paper adopts EMD to analyze helicopter acoustic signal to realize targetclassification and identification. Based on analysis and modeling of helicopter acousticsignal,its main features is the spectrum of fundamental and harmonic components. AlsoEMD algorithm is proved to have better perfromance than Fourier transform in short-time signal analysis.Because it is hard to extract the harmonic set of helicopter acoustic signal bya direct application of the Fourier transform, this paper adopts the EMD to decomposedifferent harmonic components into different intrinsic mode functions (IMF).Accordingly tothe characteristics ofhelicopter acoustic signal spectrum, a new peak frequency definition:“distance-attenuationdefinition”is given to pick out harmonic sets. Using Fourier transformof the EMD decomposition results, then the harmonic sets are extracted by criterion by peakfrequency definition.easier target identification and classification. The results show that theharmonic sets can be abstracted adaptively to realize target classification.Blasting seismic signal is also a kind of typical nonlinear signal, based on analysis of itscharacteristic, this paper adopt EMD algorithm to process measured blasting seismic signaland to study the characteristics of its speed and millisecond time, which provides a new wayfor seismic signal analysis.This paper provides a useful theoretical basis for the improvement of EMD algorithm,and expands the application fields of EMD algorithm.