FAST INITIALIZATION PROCEDURE OF THE ALTERNATING PROJECTION ALGORITHM FOR MAXIMUM-LIKELIHOOD LOCALIZATION OF MULTIPLE SOURCES

Abstract

We present an improved initialization procedure for the alternating projection (AP) algorithm which is an efficient iterative algorithm for computing the deterministic maximum likelihood (ML) estimator of the locations of multiple sources in passive sensor arrays. By utilizing the high resolution property of the sequential MUSIC (MUltiple SIgnal Classification) algorithm based on the sequential estimation technique, the procedure provides fast initial estimates that reduce significantly the number of iterations to convergence. Also these initial estimates improve greatly the possibility of global convergence. Also, we give computer simulation results to compare the AP algorithm using the proposed initialization procedure and the original AP algorithm in terms of the estimation performance and convergence behaviors.