Performance study of high resolution algorithms in sonar signal processing
2012-10-08T13:55:15Z (GMT) by
In the last two decades or so there has been great interest in the problem of estimating signal parameters from the measurements at sensor array outputs. The most important parameters are probably the directions-of-arrival (DOAs) at the array from radiating sources in the observed spatial field. This thesis is devoted to the study of algorithms and techniques which have been suggested from different points of view for the same direction estimation problem. Several classes of algorithms are examined which include the conventional beamforming methods, eigenstructure based algorithms, subspace rotation methods, decompositions techniques, and the more recently proposed weighted subspace fitting methods. The research in this thesis contains three main aspects addressing theoretical analyses, computer simulations, and practical experiments respectively. A set of simulation programs has been developed to evaluate the performance in various scenarios, and Monte Carlo tests have been carried out to support theoretical analyses. The simulation work was carried out on an IBM PC, and the computer language used was MATLAB (Matrix Laboratory), a package especially developed for matrix computations. A sonar system available in the sonar research group at Loughborough University of Technology (LUT) was modified and used to collect real data for off-line processing so as to demonstrate the algorithm performance in real experimental environments. Two scenarios were examined when the system worked in passive and active modes respectively. In the passive working mode, two emitters were employed to give uncorrelated or strongly correlated signals by using the same or different working frequencies. When working in the active mode, a single sensor was placed on the top of the receiving array which received reflections from two targets in the distance. The data was captured and then saved on floppy disks from the measurement system and processed on PCs. A large number of results are presented, analysed, and summarized in the thesis, including· both computer simulations and practical measurements. This provides a fundamental ground for further work in this and related areas.