Efficient computational algorithms of maximum likelihood estimator for localizing multiple radiating sources

Abstract

The main objective of this dissertation work is to reduce the computational complexity of the AP algorithm which is an effective algorithm for the maximum likelihood localization of radiating sources by a passive array of sensors. Since the AP algorithm is iterative in nature, there ar two ways for achieving this purpose. One is to reduce the number of iterations to convergence. The other is to reduce the computational complexity per iteration for the main iteration process and the computational complexity per stop for the initialization procedures. First, we investigate an improved initialization procedure that reduces the number of iterations to convergence and also guarantees a global convergence. For this purpose, we first develop the sequential estimation approach that lowers the detection threshold of the MUSIC algorithm noticeably through the effective removal of spatial interferences among sources. Computer simulation is done to demonstrate the detection and estimation performance of the proposed MUSIC algorithm, and also to compare the results with those of the conventional algorithm. By modifying the algorithm slightly, we also derive the improved initialization procedure that reduces the number of iterations to convergence and can also guarantees a global convergence. Computer simulation results are included to compare the AP algorithm using the improved initialization procedure with the original AP algorithm in the estimation performance and in the convergence behavior. Further, we develop a recursive computational procedure (RCP) for the sequential MUSIC algorithm and for improved initialization. The procedure reduces significantly the computational complexity per step of the proposed algorithm and the proposed initialization procedure, by transforming the computation of Hermitian forms into that of inner products of vectors only. Second, we study three different algorithms that can reduce significantly the computational complexity per iteration of...