Please use this identifier to cite or link to this item:
|Title:||Nonlinear Array Processing Techniques with Applications to Correlated Multipath|
|Authors:||Reilly, James P.|
|Keywords:||Electrical and Electronics;Electrical and Electronics|
|Abstract:||<p>The estimation of the direction of a plane wave incident upon a linear receiving antenna array and field-mapping techniques are considered in this thesis. The emphasis of the presentation is directed towards radar, and specific attention is given to the situation where the incident plane wave is corrupted by the effects of multipath propagation.</p> <p>First, the phenomenon of multipath and its experimental simulation are discussed. It is then pointed out how conventional linear array processing techniques fail in the presence of multipath propagation. These considerations lead us to consider other nonlinear array processing techniques.</p> <p>There are two such approaches considered. The first is the modification of statistical time series analysis to suit the array processing application. The Burg method and the least-squares (LS) algorithm developed by Ulrych and Clayton are two time-series methods which are discussed in detail. Results show that the Burg method is not applicable in this application, where it is shown that the LS algorithm behaves well at high SNR.</p> <p>The second approach considered is the maximum likelihood (ML) formulation. There are also two realizations of this approach which are discussed in depth. Both realizations are based on setting up the appropriate likelihood functions for the situation considered; then, the resulting structure is modified so that the required optimizations need only be performed in the specific parameter(s) of interest. This results in a more computationally efficient estimator.</p> <p>The first ML formulation discussed in applicable only to the specular multipath environment. In this situation, the direct and reflected signal components are symmetrically positioned in elevation about the normal of the array. This configuration results in a particularly simple estimator structure. The second ML formulation considered is the case where the individual signal components may be incident from arbitrary directions. The results indicate that both these ML methods perform better than the time-series techniques.</p> <p>The performance for each method is compared to the corresponding theoretical values. The results are checked by computer simulation and are validated by the use of an experimental multipath simulation system built during the course of this work.</p>|
|Appears in Collections:||Open Access Dissertations and Theses|
Items in MacSphere are protected by copyright, with all rights reserved, unless otherwise indicated.