Maximum Likelihood And Least-squares Broadband Source Localization In Beam-space
Xiao Liang Xu, K.M. Buckley · 2005
In this paper we discuss Maximum Likelihood (ML) and Least-Squares (LS) broadband source localization using reduced- dimensional beam-space array data. For broadband source locazation, ML and LS procedures have been proposed which are based on element space array data and a frequency domain statistical characterization of observations. We show that we can reduce observation space dimension with only a small reduction of source information within a selected sector, while filtering out sources outside the sector and thus reducing the parameter space dimension. This results in similar estimators but with a significant reduction in computation relative to element-space processing. We also present a new direct time domain broadband ML formulation which applies to both element-space and bearn-space processing and should be advantageous for applications which provide only a limited number of broadband snapshots. Through evaluation of Cramer-Rao Lower Bounds (CRLB's) and through simulation we show that this reduction in dimensions has little effect on the accuracy of estimates of locations of sources within the sector sector.