Thesis Submitted For The Degree Of Doctor Of Philosophy In Communications And Information System

The spatial spectrum method discribes the distribution of the parameters variance to the power spectrum,in order to estimate the parameters in term of the characteristics of the power spectrum, hence also is named as spatial spectrum estimation.This dissertation focuses on several aspects of estimation for signal parameters using spectrum methods,mainly includes Direction-of-Arrival and Time-of-Arrival.First of all,the fundamental priciples of the spectrum methods are introduced briefly,as well as the stutas quo and several problems.In the investigation of DOA,the spectrum methods mainly applies to the array signal processing,besides the uniform linear array,other array manifolds are also in the consideration. Unlike the uniform linear array,the unform circular array is able to provide 360°azimuthal coverage,within the mode space,it has the covariance matrix of the Toeplitz-like struture while all sources are uncorrelated,with which the noise variance can be made out directly,note that will eliminate an eigendecomposition when the ESPRIT is applied,consequently reduce the computational load.However,in the real circumstances,since the uniform array can not be disposed,the nonuniform array has also been getting more interest recently,several regular algorithms are presented in the prologue,especially the generalized ESPRIT for the non-uniform array.A fast approach for generalized ESPRIT without necessity of eigen-decomposition is designed,this approach constructs a pseudo-spectrum as of MUSIC algorithm,the peaks of that pseudo-spectrum are the DOA estimations,meanwhile the underlying computation is conducted to be decreased.The above work is based on the circumstance that the sources are uncorrelated,however, when there exists coherent sources,the signal convadance matrix will be rank reduced,due to which the sptectrum methods will get correct estimations ignored.For the uniform circular,the amount of the array within the mode space is less than within the real scenario because the smallest quantities of Bessel function are neglected,which means the the elements loss will get deterioration if the spatial smoothing methods are used.Caused by that,a method is proposed, which can estimate the uncorrelated sources and the coherent sources within the mode space in different stages,so that more sources can be estimated,which gives the compensation for the loss to some extent.Moreover,the noise variance is eliminated because of the differentiation operation, a fast implement of ESPRIT is given for reduction of computational load.The TOA is one of the most important techniques for localization,since the spectrum methods has the superresolution,lots of spectrum methods have been adopted in the estimation for TOA. Two difficult situations are often encountered in these methods,one is the time delay correlation matrix could be rank-reduced because of the relatively small smaple,the other is that the first path could be faded due to the NLOS environment or severe propagation condition in the LOS environment.Since that,the first path would have lower power than the subsequent ones,thus it could not be detected and give the inaccurate esitmates.An efficient TOA approach,which based on TLS-ESPRIT under the condition of small sample,is presented.An approach similar to spatial smoothing for array signal processing is adopted to recover full rank of convariance matrix of time delays,and the real domain approach to fasten the operation is also in the consideration.At the same time,an approach to mitigate the first time delay errors for a pair of terminal is also presented, which exploits the probability analysis.Several areas are constructed to distinct the estimations,and put the estimations of couples of samples into the distinct area,choose the area which has most estimations,get the last result through the avaraging.The simulation results show that the proposed approach provids high resolution time of arrival estimations.