Research On Improving The Performance Of The Near-field Source Passive Location

The signal source localization is one of the most important research aspects in array signal processing and has widely application in the field of military and civilian,such as radar,radio astronomy,seismic survey,the mobile communication and so on.Its purpose is to deal with the data received form the antenna array,to reduce the interfering signal and noise signal and to estimate the spatial location information of the interested sources.According to the different characteristics of the signal,the signal can be classified into several categories:in which according to the distance between the source and the receiver array,the sources can be classified into far-field sources and near-field ones.In this paper,the research is focused on the estimation of the near-field sources.Over the past few decades,the sources localization research is mostly focused on the estimation of the far-field sources,this assume can greatly simplify the research of spatial spectrum estimation theory,however,the theory research of the near-field source parameter estimation is relatively small in the meanwhile.The near-field sources are widespread in the real environment,such as speech enhancement,microphones,electronic reconnaissance,the seismic survey,the orientation of ultrasonic and infrared,etc.Thus the theoretical study of the near-field source parameter estimation as a new research focusing on the array signal processing,has gradually attracted the attention of scholars.After over one decade developing,many scholars have studied the Parameters Estimation Problems of near-field and have achieved many excellent achievements.Nevertheless,there still exist many problems to be solved in some aspects,such as large computation complexity,the loss of array aperture,the estimation precision,the impulse noise,mixed source coexist and so on.In this paper,we focus on the problem of the near-field sources above,aiming at exploiting the array aperture efficiently,and improving the estimation performance.The innovation of this dissertation focuses on four aspects as follows.Firstly,in order to reduce the large computing complexity of the estimation of the near-field sources,a new computational efficient method for near-field sources is proposed in this paper.On one hand,the polynomial is constructed through the symmetrical array structure and the roots of the polynomial obtain the information of angle instead of spectrum searching.The ideal Compressed MUSIC is applied to the estimation of the range information of the near-field sources.More specific,the Fresnel range is first divided into several sub-intervals.Then intersection information of the noise sub-spaces can be extracted with these available sub-intervals.Finally the new spectral function can be established based on the obtained intersection information.By doing so,the original whole-range searching is substituted by sub-space searching.Therefore it speeds up the processing time.Secondly,we propose a new array model which is similar to NEST array model for estimated losses array aperture problem existing for the near-field sources parameter estimation.Specificlly,the array aperture is enlarged by introducing the forth-order cumulate.This therefore leads to increasing number of sources which can be estimated.In addition,to avoid eigenvalue decomposition and spectrum peak search,the ideal of characteristic equation-based method(CEM)is combined with the proposed algorithm,which can greatly decreases the complexity of computation.Simulation results prove that this algorithm has great ability in expanding the array aperture.In the meanwhile,based on the main factors which influence the performance of this algorithm is the estimation precision of the high-order cumulant,and the estimation precision of the high order cumulant is different,the new algorithm adopts to the weighted factor to optimize the algorithm.Simulation results show that the optimization algorithm can improve the performance without increasing the computation complexity.In addition,we consider the estimation of the near-field sources in impulsive noise.In order to improve the disadvantage of the traditional algorithms for the impulsive estimation,we proposes the lp-norm of the residual fitting error matrix as the objective function for subspace decomposition,adopts the alternating convex optimization(ACO)to solve the resulting non-convex matrix decomposition under the lp-norm minimization framework and replace the non-convex optimization problem of the matrix by the convex optimization on problem of multidimensional vector in order to reduce the complexity of computation.Simulation results show that the proposed algorithm has better performance and more effective in strong impulsive noise than the algorithm based on the fractional lower order statistics.Finally,we research on the parameters estimation algorithms for coexisted far-field sources and near-field sources then propose a new algorithm based on the oblique technology,in which the sources are estimated at two different stages.The DOA of far-field sources can be obtained through MUSIC(multiple signal classification)algorithm.Then sensor noise power is estimated and eliminated from the covariance in order to improve the accuracy of the estimation of near-field sources.The contributions of far-field sources can be eliminated from the covariance matrix by oblique projection technology,while only the information of near-field sources is reserved and to extend A Sparse Covariance-Based Estimation Method to the parameters estimation of the near-field sources.The proposed algorithm has better performance and higher estimation precision,and includes the ability to solve the two closed-spaced sources with high resolution and accuracy.