Blind Source Separation Based Multidimensional Parameters Estimation Of Array Signals

The estimation of arrival parameters with array sensors is an important research feldof array signal processing, it has found wide applications in a variety of felds, such asradar, communications, sonar, seismology, etc. So, its study is of great theoretical andpractical value. Based on the theory and algorithms of blind source separation, this dis-sertation aims at the development of high-resolution and robust arrival parameters esti-mation algorithms, involving direction-of-arrival (DOA) and range in diferent situations,i.e., DOA estimation of far-feld sources, DOA estimation of far-feld sources in the pres-ence of unknown mutual coupling, joint DOA and range estimation of near-feld sources,DOA estimation of far-feld coherent sources. Computer simulations are also carried outto verify these algorithms. The main work of the dissertation can be summarized as fol-lows:1. Basic principles of blind source separation (BSS), especially independent com-ponent analysis (ICA), are frst introduced. According to the constructing criteria of costfunction, ICAalgorithmsareclassifedintotwocategories. TwoclassicalICAalgorithms,which are referred as FastICA (Fast Independent Component Analysis) and JADE (jointapproximate diagonalization of eigenmatrices), are studied in detail. A comparison be-tween BSS and arrival parameters estimation with array sensors is carried out to investi-gate the feasibility of exploiting BSS to estimate arrival parameters with array sensors indiferent practical environments.2. DOA estimation of far-feld narrowband sources is frst studied for uniform lineararray(ULA),uniformcirculararray(UCA)andL-shapedarray. BasedonBSSalgorithms,corresponding algorithms are proposed for these array confgurations. Besides spatial in-formationinarrayconfgurations,BSS/ICAbasedDOAestimationalgorithmsalsoexploithigh-order statistics or temporal structure of sources. Unlike most conventional high-resolution algorithms, BSS/ICA based DOA estimation algorithms obtain DOAs sepa-rately, thus the number of array elements can be reduced. In dealing with very closelyspaced emitters, BSS/ICA based algorithms is superior to conventional high-resolutionmethods.3. DOA estimation in the presence of mutual coupling is studied. Based on BSSand models for mutual coupling matrix, two algorithms are proposed for ULA and UCA, respectively. In the frst stage, BSS algorithms are applied to estimated the generalizedarraymanifoldmatrix. Inthesecondstage,basedonthebandedsymmetricToeplitzmodelfor the mutual coupling of a ULA and the circulate and symmetric Toeplitz model for themutual coupling of a UCA, the problem of joint estimation of DOAs and mutual couplingcoefcientsisthenconvertedtosolvemultipleseparablenonlinearleastsquaresproblems,whereDOAsareobtainedbyfndingthemaximumvaluesofspatialspectrums. Comparedwith most existing algorithms, to obtain DOA estimation, the proposed algorithms doesnot need to estimate mutual coupling coefcients, and DOAs are obtained by searchingglobalmaximumvaluesofmultiplespatialspectrumsinsteadofpeaksearchingofaspatialspectrum, thus avoiding converging to local extremums.4. Joint range and DOA estimation of near-feld sources is studied, BSS based al-gorithms are proposed using ULA and UCA, respectively. For ULA, the array manifoldmatrix is frst estimated by means of BSS algorithms. Initial estimation of range and DOAis then obtained using Fresnel approximation. To alleviate the infuence of estimation biasintroducedbyFresnelapproximation,nonlinearleastsquaresalgorithmsareappliedtoob-tain more accurate estimation. The proposed algorithm does not need parameter pairingoperation and can resolve more sources with the same number of sensors. For UCA, thearraymanifoldmatrixisalsofrstestimatedusingBSSalgorithms,then,estimationsofaz-imuth angles are estimated, according to the conclusion that the actual near-feld azimuthangle is equal to the estimated far-feld angle, regardless of the range and elevation angle.Finally, joint range-azimuth-elevation estimation is obtained using Fresnel approximationand nonlinear least squares.5. Based on the rotational invariance property of ULA, the problem of DOA estima-tion of coherent sources is studied, and a fast DOA estimation algorithm without eigende-composition or peak searching is proposed. The proposed fast algorithm fully exploits theshift invariance property of ULA, only covariance matrix of array output is adopted, andclosed-form solutions of DOAs are obtained through polynomial rooting. Combined withphase mode excitation technique, the proposed fast algorithm can also be applied to1-DDOA estimation of coherent sources using UCA. Based on the proposed fast algorithmand BSS, a novel algorithm for DOA estimation of multiple groups of coherent sourcesis proposed, the rotational invariance property of ULA is again adopted to eliminate theinherent permutation indeterminacy introduced by BSS algorithms. In the last chapter, the work in this dissertation is summarized and the contents forfurther study are also pointed out.