4 Localization of sources in a sector: Algorithms and statistical analysis

Abstract

Summary In this chapter we have addressed eigenspace spatial-spectrum based source location estimation. Specifically, we have considered estimation of source locations which are restricted to a sector within the array field-of-view. Several algorithms have been studied and compared, using resolvability and estimate accuracy as performance criteria. Analysis has been based on statistical expressions of estimate variance and bias, which have been derived recently for eigenspace spatial-spectrum based estimators. This study concentrated on two approaches to enhancing estimation of source locations in a location sector — beam-space preprocessing, and sector focusing via the CLOSEST approach. Results indicate that both approaches improve resolution and reduce location estimate bias relative to popular algorithms such as MUSIC and MIN-NORM (which do not focus on a location sector), while providing estimate variance comparable to MUSIC (which provides lowest variance of the general weighted MUSIC class of element-space estimators). Algorithms were considered which are applicable to arbitrarily configured arrays. Narrow-band element-space and beam-space processing was addressed explicitly, although discussions and results are applicable to broad-band problems assuming that appropriate broad-band preprocessing is employed.