Study On Multidimensional Parameter Estimation For Near-field Sources

Multidimensional parameter estimation of near-field sources are studied in this thesis.The details of this thesis are organized as follows.1. Three novel methods for joint frequency DOA and range estimation of near-field sources are presented. Compared with coventional methods available, the proposed methods can efficiently improve the performance of the parameter estimation owing to decrease the loss of the array aperture. Furthermore, the parameters can be paired automatically. The first two methods are based on four order cumulant, however, the first method need to costruct three four-order cumulant matrices and a high matrix, so it has high computational load. The second method only need to costruct two four-order cumulant matrices and need not to costruct high matrix, so it computate efficiently. The third method is based on propagator method. It avoids the estimation and eigen value decomposition of the covariance matrix(or four-order cumulant matrix) of the received signals. So the computational complexity of the proposed algorithm is reduced greatly with the maintenance of good performance for parameter estimation. Finally, simulation results demonstrate the effectiveness of the proposed methods.2. Two new methods for four-dimensional parameter estimation of near-field sources are presented. The first method is based on cumulant, Frist, three four-order cumulant matrices are constructed, and then two new matrices have been obtained from the transformation of these matrices. The four parameters are directly given by the eigenvalues and eigenvactors of the new matrix. Compared with several existing methods, the proposed method can efficiently avoid the loss of array aperture, so the performance of parameter estimation is improved , and it can achieve automatic pairing. The second method is based on second statistics and use full symmertric cross-array. Because of the additive noises of the x-axis subarray and y-axis subarray are uncorrelated with each other, we can use this property to contrust cross-correlation matrix to elimate the effect of noise. it only need one process for parameter pairing. Also it avoids using four-order cumulant so it is computationally efficient. Finally, the simulation results are presented to validate the performance of the proposed methods.