Joint Multi-dimensional Parameter Estimation Of Near-field Sources Based On High-resolution Algorithm

Parameter estimation,an important branch of array signal processing,which has been widely used in the fields of national economy,scientific research and military defense,such as communication,radar,sonar,seismic survey,radio astronomy and biomedicine,has great research significance.Most parameter estimation algorithms assume that the radiating sources are located in the far-field of the antenna array,but in many practical applications,there may be cases that the sources are located in the near-field,or that the far-field sources coexist with the near-field sources.This dissertation mainly discusses the parameter estimation problems of near-field sources and mixed sources.It further researches the problems of high computational complexity,array aperture loss,poor estimation accuracy and parameter matching in the existing algorithms and proposes feasible solutions.The main content of this dissertation can be summarized into the following four parts:In the first part,the theory and mathematical foundation of near-field parameter estimation are introduced.Firstly,the relevant basic mathematical knowledge is introduced.Matrix algebra and high order cumulants are commonly used in near-field parameters estimation.Secondly,the signal model of near-field source received by uniform linear array is established.Then the Cramer-Rao bound for performance analysis of near-field parameter estimation,is derived.In the end,we summarize the three approaches of the existing nearfield source estimation algorithms and introduce two classical near-field estimation algorithms in detail: the near-field rooting algorithm and the esprit-like algorithm based on high order cumulants.The second part studies the basic concept of subspace algorithm-array manifold,which contains all parameter information of near-field source.The study of array manifold characteristics is of great significance to analyze the performance of array parameter estimation and to propose a new algorithm for parameter estimation.Firstly,in view of the spherical wave propagation characteristics of near-field source signals,the mathematical model of manifold of a N-elements linear array receiving near-field signals is constructed.It can be seen that the near-field array manifold is a surface embedded in N-dimensional complex space.And the far-field array manifold is compared with the near-field array manifold.Secondly,the local characteristics of near-field array manifold are studied by using differential geometry.By establishing the sliding coordinate system,the mathematical expressions of the near-field array manifold arc length and the first-order curvature are given.The precision and resolution performance of near-field parameter estimation are analyzed by using these two local characteristic parameters.Finally,the effect of array manifold on near-field beamforming is analyzed.A formation optimization algorithm for near-field beam synthesis is proposed.In the case of fixed array number and spacing constraints,differential evolution algorithm is used to optimize the position of the element,so as to reduce the peak sidelobe level of the beam pattern.In the third part,we study the parameters estimation algorithm of near-field non-circular signal.Firstly,the "circle" and "non-circle" characteristics of signals are introduced.It's pointed out that the existing algorithm does not make full use of the information of the elliptic covariance matrix of the signal.Then the received signal model is established when the near-field source is non-circular signal.Secondly,a real-value multi-dimensional parameter estimation algorithm for near-field non-circular signals is proposed based on dualpolarization sensor array.By taking use of the symmetric property of the array and the noncircular characteristics of the signal,the angle,distance and polarization parameters are decoupled from the array manifold vector.According to the rank reduction theorem,the high-dimensional spectral peak search problem of traditional subspace algorithm is simplified into multiple one-dimensional searches.Then the computational complexity can be effectively reduced as well as by the real value operation.In addition,by making full use of the non-circular characteristic of the signal,the maximum resolvable source number and parameter estimation accuracy are improved.Finally,a fast parameter estimation algorithm for near-field non-circular signals is proposed.On the basis of decoupling the near-field steering vector,the traditional spectral peak search is replaced by polynomial rooting.Once the array formation structure is determined,the position parameters of the signal source can be obtained simply by solving the polynomial.It's easy to implement the algorithm without any additional operation such as parameter matching.In the fourth part,the multi-objective parameter estimation algorithm is studied when the far-field source coexists with the near-field source.Both far-field and near-field sources may coexist in the situations such as microphone array based sound source positioning,underwater sonar system,indoor navigation system and electronic countermeasures system.Pure far field source and pure near field source can be regarded as two special cases of mixed far-field and near-field sources.Firstly,the possibility of using the far-field or near-field parameter estimation algorithm to process the mixed sources data is analyzed.This may lead to performance degradation or even invalidation of parameter estimation.A novel algorithm based on sparse subarrays is presented to solve the mixed sources localization problem.Through the selection of specific array elements,the fourth-order cumulant matrix with only angle information is constructed.Using the dual-size shift invariance in and between subarrays,the fuzzy free rough estimation and fuzzy fine estimation of angles are obtained respectively.The fine DOA estimation with ambiguity and coarse DOA estimation without ambiguity are obtained respectively.The fine estimation of mixed source is obtained by disambiguation.The estimated DOAs of mixed source is taken into another constructed cumulant matrix constructed,and the near-field source range estimation is obtained by onedimensional search.Finally,a three dimensional parameters estimation algorithm for mixed sources is proposed based on cross array.Firstly,the elevation angles of mixed sources and the distance parameters of near field sources are estimated by using the received data of zaxis array.Substituting the estimated elevation angles and ranges parameter into the received data of x-axis array,the azimuth estimation of mixed sources is obtained by one-dimensional azimuth search.