Research Of Near-field Source Localization Method Based On Higher-order Cumulant

As an important problem in array signal processing, source localization thatoriginated from the old direction finding studies which direction of the receivedsignal comes from. It contains far-field source localization (also called DOA) andnear-field source localization. The early studies of source localization mainlyfocuse on the far-field scenario with planar wavefronts assumption, hence thelocalizations are only decided by the parameters of elevation and azimuth. Butwhen the sources are close to the array (viz., in the near-field case), the far-fieldassumption is no longer valid, the spherical wavefronts assumption must be usedto describe the near-field sources, and range parameters are added to characterizethe localizations. Also near-field source localization is widely applied in radar,sonar, speech enhancement, electronic surveillance and seismic exploration etc.For the limitation of the mathematics means, traditional near-field sourcelocalization methods can suppress only stationary white Gaussian noise and haveno way in eliminating colored Gaussian noise or non-Gaussian noise. With thedevelopment of higher-order statistics (HOS), the methods based on HOS canallow stationary colored Guassian noise or symmetry distribution non-Guassiannoise. The method based on second-order cyclic statistics can suppress stationarynoise and interference with different cyclic frequency, but it can only estimateazimuth parameters of the sources, so it needs the further research. Because of theadvantages of using HOS, we choose fourth-order cumulant as the mainmathematics method to solve the near-field source localization problem.The main research of this dissertation focuses on the near-field sourcelocalization problem. Four new source localization methods based on fourth-ordercumulant are proposed, which combine the source number detection method withthe DOA matrix method well. These new methods can solve the source numberestimation and parameter pairing problems in the 2-D (dimension) planar and 3-Dspatial near-field source localization.This dissertation consists of six chapters. The basic knowledge and theexisting near-field source localization methods are introduced in the first threechapters;the auther's main research work is introduced in the fourth chapter andfifth chapter;and the last chapter gives the conclusion of the whole dissertation.In chapter one, the research background, development history and presentstatus of near-field source localization are summarized, and the research contentof this dissertation is proposed.In chapter two, the fundamental mathematics knowledge involved in thisdissertation is introduced, basic assumption of the near-field source localizationmodel and several kinds of typical mathematics models are presented. These arethe foundation for the introduction of classical methods and the new proposedmethods of this dissertation.In chapter three, several classical near-field source localization methods areintroduced. The estimation performances of the localization methods in thedifferent noise and SNR environments by computer simulations are compared,and the advantages and disadvantages of the methods are analysed.In chapter four, for the disadvantages of the classical methods, by using themodified AIC criterion to estimate the near-field source number, two new 2-Dplanar near-field source localization methods based on higher-order cumulant areproposed in the case of carrier frequencies are known and unknown. The newmethods solve the problems of source number estimation and parameter pairing.In chapter five, based on the new methods in chapter four, two new 3-Dspatial near-field source localization methods based on higher-order cumulant andthe cross array are proposed in the case of carrier frequency parameters are knownand unknown. These two new methods also solve the problems of source numberestimation and parameter pairing in 3-D spatial near-field source localization.In chapter six, a brief summary of the dissertation is given. The suggestionfor future research related to near-field source localization is put forward.The main contributions of this dissertation are listed as follow(1) For the most existing methods can't detect source number, the modifiedAIC criterion which was used to estimate the far-field source number is applied toestimate the near-field source number. Theoretical analysis and some computersimulations show that the criterion is an efficient near-field source numberdetection method in the additive Guassian noise environments. It can inerrablyestimate the source number under a certain SNR and snapshots condition.(2) For 2-D planar near-field source localization problem, a new methodbased on fourth-order cumulant and DOA matrix method for joint azimuth-rangeestimation of sources is proposed in the case of carrier frequencies are known. Itcan accurately estimate source number, doesn't need spectral peak search, higher-dimension cumulant matrix or parameter pairing, and its computer burden ismoderate. Based on this new method, another new method for joint carrierfrequency-azimuth-range estimation of sources is presented in the case of carrierfrequencies are unknown, its efficiency is also verified by computer simulations.(3) For 3-D spatial near-field source localization problem, based on fourth-order cumulant and the cross array, a new method for joint elevation-azimuth-range estimation of sources and another new method for joint carrier frequency-elevation-azimuth-range estimation of sources are proposed in the case of carrierfrequencies are known and unknown. These two new methods also can accuratelyestimate source number, don't require spectral peak search or higher-dimensionmatrix, needs only simple 1-D parameter pairing, and their computer burden ismoderate. Additional, the research about 3-D spatial near-field source localizationproblem with carrier frequencies unknown isn't found until now.And this dissertation has some groping actions to the three questions ofnear-field source localization as follow(1) Source number detection in near-field source localization.(2) Source parameter pairing in near-field source localization.(3) Higher-dimension parameter estimation in near-field source localization.