DOA estimator performance assessment in the pre-asymptotic domain using the likelihood principle

• Authors: Ben A. Johnson;Yuri I. Abramovich
• Venue: Signal Processing
• Year: 2010

Performance assessment of algorithms for direction of arrival (DOA) estimation are typically done using large-sample justified asymptotic constructs such as consistency, efficiency, and the Cramer-Rao lower bound. The performance in parameter accuracy (usually the mean square error of the DOA estimate) of the algorithm relative to the true parameters of the sources is evaluated to determine if the algorithm is accurate, robust, computationally efficient, etc. However, performance assessment of the algorithm in practical circumstances with limited data sample volume cannot use these methods, because asymptotic statistical behavior is no longer met and the true location of the sources is in general unknown. This paper reviews the application of an performance assessment technique referred to as expected likelihood in such practical small-sample circumstances, and provides simulation and real-world examples of the capabilities provided by expected likelihood which does not rely on knowledge of the true source locations. Uses of the approach in other areas such aiding of numerical optimization, model order determination, and determination of appropriate diagonal loading in LSMI applications is also reviewed.