Threshold performance for conditional and unconditional direction of arrival estimation
Yuri I. Abramovich, Ben A. Johnson · 2012
In this study, we provide comparative analysis of the so-called “threshold conditions” for conditional (deterministic, CML) and unconditional (stochastic, UML) maximum likelihood estimation. Specifically, we analyze SNR and sample support values where the accurate UML and CML DoA estimation starts to rapidly divert from the Cramér-Rao bound (CRB) due to the onset of severely erroneous DoA estimates (“outliers”). Theoretical predictions based on the recent large random matrix theory (RMT) findings are verified by Monte-Carlo simulations for sources with different levels of inter-source (spatial) or temporal correlations.