Statistical nonidentifiability of close emitters: Maximum-likelihood estimation breakdown and its GSA analysis

摘要

We investigate the ldquoambiguity regionrdquo associated with erroneous maximum-likelihood (ML) direction-of-arrival (DOA) estimates of closely spaced signals, near and below the ldquoresolution limitrdquo. We demonstrate that the general statistical analysis (GSA) technique can accurately predict the ambiguity region for a given scenario. We consider that this prediction may be used together with the Barankin bound (BB) to form a more comprehensive description of maximum-likelihood estimation (MLE) performance in the problematic ldquothreshold regionrdquo where ML techniques suffer from a dramatic failure rate (ldquoperformance breakdownrdquo).