Research On Methods Of Parameters Estimation Based On Polarization Array

With the development of array signal processing technology, polarization array is widelyused because it can exploit the incident sources’ polarization diversity. Sensors of polarizationarray are usually composed of spatially co-located antennas, separately measuring the incidentsources’ three electric-field components and three magnetic-field components. In this paper,algorithms for parameters estimation based on polarization arrays have been studied in thispassage, and characteristics of polarization array and disadvantages of existing parametersestimation algorithms have been taken into consideration. The main content is as follows:1. The parameters estimation algorithms for polarized signals based on quaternion arestudied. Existing parameters estimation methods, which mostly uses long-vector mode, do notmake full use of polarization dimension information of signals. The quaternion formalismprovides an elegant and compact way of handling multi-component data of vector sensors, andtherefore provides a more accurate estimation of the signal subspace. As far as high-dimensionsearch during the process of muiti-dimension parameters estimation on the basis of existingmethods is concerned, this paper proposes a new kind of reduction parameters estimation methodusing COLD arrays. Firstly, parameters separation of quaternion data covariance matrix iscarried out and DOAs are acquired via peak-search of a two-dimension quaternion spectralmatrix (QSM). Then, closed-form estimation of polarization parameters is realized throughconstructing polarization rotation matrix. The presented algorithm outperforms long-vectorMUSIC (LV-MUSIC) for DOA estimation to some extent, while error accumulation is eliminateddue to the independence of DOA estimation and polarization estimation. Furthermore, thepresented algorithm has a large reduction in computational effort together with a lower memoryrequirement compared with long-vector MUSIC algorithm.2. Parameters estimation methods using sparse signal processing are researched. Mostexisting subspace-based methods often suffer performance degradation under small number ofsnapshots as well as under low SNR as a result of subspace diffusion, and they can not processcoherent sources very well. Hence, sparsity of signals is utilized in this paper to lighten or avoidthese shortcomings. Firstly, this paper introduces sparse representation, sparse reconstructionmethods and sparse model of narrowband arrays. Then, an L1-SVD parameters estimationmethod is introduced, which makes use of signals sparsity and improves parameters estimationperformance under low SNR as well as with small snapshots. Finally, considering performancedegradation of L1-SVD method under wrong estimation of sources’ number, a new algorithmutilizing prominent eigenvector of data covariance matrix is proposed, which further improvesestimation performance and robustness and can process coherent signals well since it does notneed estimation of subspace.3. The research is also related to parameters estimation and tracking methods based ondistributed electromagnetic vector sensor. In the presence of moving targets, traditionalsubspace-based methods provide poor performance due to its necessity of continuous calculation of data covariance matrix and peak-search of spectrum. Moreover, as the number of arraysensors increases, their performance worsens. Though many are the advantages offered by a kindof six-component electromagnetic vector-sensor to angle and polarization estimation, such aslittle volume, good resolving power for the sources’ polarization-difference and quite simpleestimation method, what is negligible is the mutual coupling across the six collocated antennas.So, distributed electromagnetic vector sensor is applied in this passage. Fast subspace trackingmethods, uni-vector-sensor ESPRIT method and “vector cross product” method are introduced.Then, a special distributed electromagnetic vector sensor model is researched. Finally, a newparameters estimation and tracking method for multiple moving targets is presented on the basisof FDPM and extended uni-vector-sensor ESPRIT method and “vector cross product” algorithm.The algorithm provides simple structure, high accuracy, small computation complexity and usageof aperture information, and therefore has attractable application thanks to its advantage in bothhardware cost and fine performance.