Study On DOA And Polarization Parameters Estimation With Vector-Sensor Array Based On Quaternion

The electromagnetic vector-sensor is composed of various subantennas with different polarization characters.The vector-sensor array,which consists of electromagnetic vector-sensors,can obtain three-dimensional information of each incident signal,including spatial,time and polarization domains,thus the vector-sensor array outperforms than that of the conventional scalar-sensor array on weak and small target detection,adaptive beamforming,multidimensional parameters estimation,and the rest.The received data of the vector-sensor array modeled by quaternion algebra has a stronger orthogonal constraint,thus the direction of arrival(DOA)and polarization parameters can be estimated more accurately based on quaternion model.This dissertation focuses on the quaternion model based high resolution algorithms for the co-centered and stretched vector-sensor array.The main contributions in this dissertation are summarized as the following three aspects.1.Multidimensional parameters estimation algorithms for the one-dimensional co-centered vector-sensor array are studied.The conventional dimensional reduced estimation of signal classification(DRQ-MUSIC)algorithm has a problem that the polarization parameters need to be re-estimated by long-vector model.Focusing on this problem,the dimensional reduced estimation of signal parameters via rotational invariance techniques(DRQ-ESPRIT)algorithm is first proposed.The proposed algorithm points out that there exists a “quaternion model coherence” problem in the DRQ-MUSIC algorithm,resulting in the loss of the polarization information of each incident signal and the orthogonal constraint of the received data in the covariance matrix of the received data.The proposed algorithm avoids this problem by replacing the covariance matrix by cross-correlation matrix,thus the azimuth angle,elevation angle,polarization auxiliary angle and polarization phase difference of each incident signal can be estimated by utilizing the structure characters of quaternion model and the coupling relationship between polarization and angle domains.In order to solve the “quaternion model coherence” problem,the fake virtual symmetrical aperture expanded quaternion ESPRIT algorithm is then proposed.Based on the principle of quaternion model,together with the noncommutation of quaternion multiplications,the proposed algorithm solves the “quaternion model coherence” problem by modifying the multiplication sequence between the electromagnetic expression and spatial steering vector.In order to make up the loss of degrees of freedom(DOF)and aperture of the vector-sensor array caused by spatial-smooth-liked process,the Khatri-Rao subspace process is extended into the quaternion model,together with the symmetrical operation of steering vector,the proposed algorithm can estimate more signals with higher accurate.Finally,focusing on the half DOF loss of the vector-sensor array caused by the quaternion model,the 2 level nested vector-sensor array based quaternion ESPRIT algorithm is proposed.The array structure of the 2-level nested scalar-sensor array is applied into vector-sensor array to enlarge the DOF and aperture of the vector-sensor array,thus more signals can be estimated,and the accuracy of parameters estimation is improved.2.Multidimensional parameters estimation algorithms for the two-dimensional co-centered vector-sensor array are studied.Focusing on the L-shaped vector-sensor array which consists of half-wavelength spaced co-centered orthogonal loop and dipole(COLD)sensors,an L-shaped array based DRQ-MUSIC algorithm is first proposed.According to the proposed algorithm,the search steering vector is divided into three parts by utilizing the geometry of the L-shaped vector-sensor array.Based on the spatial-smooth-liked process,a full column rank matrix is constructed by diagonal elements of the cross-correlation matrix.Each part of the search steering vector is estimated separately,thus the high dimensional parameters search problem is simplified to a one-dimensional parameter search problem,the computational complexity is reduced considerably.In order to enlarge the aperture of the vector-sensor array and make the array structure more flexible,the L-shaped vector-sensor array is extend to an two-dimensional vector-sensor array which is composed of one half-wavelength spaced uniform linear subarray along the z-axis,and one sparsely and arbitrarily spaced planar subarray along the x-y plane,then the polarization diversity based Q-ESPRIT algorithm is proposed.According to the proposed algorithm,the array manifold is derived by constructing the rotational invariance structure,together with the coupling relationship between the polarization and angle domains brought by the added dipole in the reference point,the DOA and polarization parameters can be estimated unambiguously.3.In order to make full use of the spatial information of each component of the electromagnetic vector-sensor,multidimensional parameters estimation algorithms for the stretched vector-sensor array are studied.The L-shaped array which consists of the stretched COLD sensors is first studied,both dipole components and loop components are half-wavelength spaced.In order to avoid the four-dimensional spectral peak search of the conventional Q-MUSIC algorithm,the generalized rotational invariance based dimensional reduced MUSIC algorithm is proposed.Based on the generalized rotational invariance structure between the two subarrays,together with two dimensional reduction processes,the high dimensional parameters search problem is simplified to a one-dimensional parameter search problem,the computational complexity is reduced considerably.Furthermore,sparsely stretched COLD sensors are studied to enlarge the aperture of the vector-sensor array,the loop components are still half-wavelength spaced,while the dipole components are sparsely and uniformly spaced,and the quaternion based generalized rotational invariant dimensional reduced MUSIC algorithm is proposed.Adding a dipole in the reference point,the proposed algorithm uses the quaternion model and the coupling relationship between polarization and angle domains to solve the ambiguity problem caused by the sparse stretching process.Finally,the sparsely stretched COLD sensors are extended into the sparsely stretched three-component electromagnetic vector-sensors,and the generalized rotation invariance based quaternion MUSIC is proposed.The accuracy of DOA and polarization parameters is improved with quaternion model,and the ambiguity problem caused by sparsely stretched dual-dipole components is solved by the polarization diversity of three-component electromagnetic vector-sensors.