Investigation On Estimation Of Signal Parameters Using Vector Sensor Arrays

Array signal processing processing technique has played a fundamental role in many applications involving radar,sonar and wireless communications.Customary array signal processing technique uses scalar sensors,each of which can only measure the one-dimensional scalar quantities of the wave field induced at the sensor.Recently,it is found that exploiting an array of "vector sensor" can provide measurements of more than one scalar quantities of the wave field,therefore,an array of vector sen-sors can provide more superior systemic performance that that offered by an array of scalar sensors.Therefore,it is of great significance to develop vector sensor array sig-nal processing methods.Following the works conducted by previous researchers,this dissertation investigates the signal processing methods using acoustic/electromagnetic vector sensor arrays,and proposes several new vector sensor array parameter estima-tion methods.The main innovations of the work are as follows.Part I:Research on signal processing methods with acoustic vector sensor(ACVS)arrays1)We consider joint angle and delay estimation for underwater acoustic Multicar-rier CDMA systems using a vector sensor.We first incorporate the temporal dimension to the vector sensor response vector to construct the temporal-vector-sensor(TEVES)response vector for the MC-CDMA signals.Then,the subcarrier information is used to decorrelate the signal coherency and a MUSIC-based algorithm is proposed for JADE of a "desired source" operating in the multipath environment.The proposed algorithm is in a tree structure,containing two ?-MUSIC steps for estimation of delays,two ?-MUSIC steps for estimation of elevation angles and one ?-MUSIC step for estimation of azimuth angles.In addition,a closed-form ESPRIT-based algorithm is presented for the scenario where the multipath propagation signals are emanated from a single source.The Cramer-Rao bound(CRB)for the problem under consideration is also provided.2)We develop a method for two-dimensional direction finding of coherent signals with a linear array of acoustic vector sensors.We first formulates a PARAFAC model by using the spatial signature of vector hydrophone array to extract vector sensor array manifolds by PARAFAC analysis,without requiring to perform spatial smoothing or vector-field smoothing to decorrelate the signal coherency.We also establish that the 2D directions of K coherent signals can be uniquely determined by PARAFAC analysis,as long as the number of sensors L ? 2K-1.In addition,because the acoustic vector sensor array manifold does not contain time-delay phase factor,the proposed algorithm may not suffer direction cyclical ambiguity when the spacing between sensors extends beyond a half-wavelength.The inherent structure of acoustic vector sensor also allows the array aperture extension to offer enhanced angle estimation precision.Part II:Research on signal processing methods with electromagnetic vector sensor(EMVS)arrays1)Two methods for DOA estimation of partially polarized/mixed(simultaneous-ly existing of both completely polarized and partially polarized)signals with dual-component vector sensor array are presented.For the first algorithm,we employ a uniformly linear array,where each sensor having two single-polarized elements.By exploiting the array geometry and its shift invariance property,we construct a set of data correlation sequences using array output and its conjugate to transform the problem of the direction finding to that of the estimation of the complex sinusoid fre-quencies.The rank-2 signal correlation matrices of the PP signals become real-valued amplitudes of the complex sinusoids,and thus,the rank-2K signal subspace of the array output transforms to the rank-K signal subspace of the constructed correlation sequences.We then apply the conventional MUSIC and ESPRIT methods to the cor-relation sequences to estimate the directions of the signals.We also show that the directions can be uniquely determined as long as the number of sensors is greater than that of the PP signals.The proposed method is also valid for scenarios,where both CP and PP signals coexist.Moreover,it is "blind" in that it does not require any prior information on the degree of polarization of the signal.For the second algorithm,we develop subspace-based methods for DOA estimation using an array of cross-dipoles,for the scenario where both completely polarized(CP)and partially polarized(PP)electromagnetic(EM)signals coexist.Three typical algorithms:MUSIC,ESPRIT and generalized ESPRIT algorithms are derived.The MUSIC-based algorithm constructs two estimators,which are termed as CP estimator and PP estimator,to extract the DOA's of CP signals and PP signals separately.These two estimators can be combined to discriminate the PP signals from the CP signals without requiring the estimation of the degree of polarization(DOP)of the signals.Moveover,the CP estimator is able to estimate the polarizations of the CP signals.The ESPRIT-based algorithm finds the DOA estimates of the CP and PP signals simultaneously from the ESPRIT's eigenvalues.It can also be used to signal classification.The generalized ESPRIT-based algorithm yields the DOA estimates of the CP and PP signals simultaneously by performing an one-dimensional(1D)search,but it is not applicable to differentiate the CP signals from the PP ones.Furthermore,the ESPRIT and generalized ESPRIT based algorithms cannot estimate the polarizations of the CP signals.2)Two direction finding methods using six-component electromagnetic vector sen-sors are developed.For the first algorithm,by exploiting the temporal invariance of the monochromatic signals,we relate multiple time-delayed data sets to the parallel faction(PARAFAC)model.Furthermore,we establish that as long as the number of data samples N ? 2K-1 or N? 4K-1,K uncorrelated monochromatic completely polarized or partially polarized signals can be uniquely resolved by PARAFAC decom-position.This identifiability result is more general than that in[134],as it does inot rely on the linear independence of the EMVS steering vectors.For the second algorith-in we firstly apply the temporally smoothing technique to improve the identifiability limit of a single vector.In particular,we establish sufficient conditions for constructing temporally smoothed matrices to resolve K>2 partially polarized monochromatic sig-nals with a single vector sensor.Then,we derive an efficient ESPRIT-based method,which does not require any calibration signals or iterative operations,to jointly esti-mate the azimuth-elevation angles and the mutual coupling coefficients.Finally,we present CRB for the problem under consideration.The performance of proposed algorithms is verified through various computer simulations.