Application Of HHT And SVM In The 4-UCA Source Number Estimation

As an important research direction of array signal processing,spatial spectrum estimation technology has been widely used in many fields such as radar communication,electronic countermeasure,biomedical detection,seismic signal monitoring and so on.The main purpose of spatial spectrum estimation is to estimate the number of signal sources and estimate the direction of arrival of the signal source.The accurate estimation of the number of signal sources is one of the preconditions for many high resolution DOA estimation algorithm.If the number of sources is not estimated,the accuracy of high resolution DOA estimation will be greatly affected.Therefore,the estimation of signal source number is one of the basic tasks of spatial spectrum estimation.At present,the method of estimating the number of signal sources can be classified into two categories: one is the traditional signal source number estimation method based on discriminant function,and the other is pattern recognition classification based on machine learning.The main work of this paper is to reseach the estimation of the number of signal sources in a determined case by using the four array uniform circular array(4-UCA)based on Hilbert-Huang Transform(HHT)and Support Vector Machine(SVM).Firstly,this paper introduces the model of 4-UCA and the date model of each array element.Analysis the influence of the accurate estimation of the number of signal sources on the estimation of DOA by simulation experiment.Next analysis the similarities and differences between the traditional source estimation algorithm and the algorithm based on machine learning.Secondly,this paper expounds the theories of HHT and SVM.For HHT,this paper introduces its basic principle,two core algorithms,the analysis process.For SVM,this paper summarizes its therotical basis and classification principles.Aiming at the signal source number estimation,because of the equivalent relation between the instantaneous phase of HHT and the original signal,this paper propose a feature extraction method based on instantaneous phase of HHT after analyzing the phase difference of received signal between the uniform circular channels.So the instantaneous phase of the first 4 IMF components of the signal is obtained as a feature to distinguish the signal with different number of sources.Since the difference of instantaneous phase between channels is a variable,therefore,the correlation coefficient is introduced to measure the correlation degree between the two channels the instantaneous phase of the corresponding IMF component.Then constructe the correlation coefficient matrix of the IMF component and get the eigenvalues of the matrix.Using the eigenvectors of the correlation matrix of 4 IMF components to constructe the feature vector.In the case of determined condition,the number of signal sources is up to be 3 when using 4-UCA to estimate the number of signal sources.a muti-class classifier on the basis of LibSVM is designed and its structure,classification algorithm and other related parameters are determined through the heoretical analysis and experimental tests.The simulation experiment and actual data test results show that the proposed method is effective and feasible.