Generalized consistent robust Capon beamforming for arbitrarily large arrays

In this paper, the consistency of sample robust Capon beamforming (RCB) solutions that are constructed under signature-mismatch constraints from a set of received array observations is revised and improved. Particular emphasis is given to the class of robust filters heuristically modeling the adverse effects of practical finite sample-size conditions as due to an imperfect knowledge of the effective spatial signature. In contrast, and as in practice, a small sample-size relative to the array dimension is identified in this paper as the actual source of filter estimation errors under unknown second-order statistics. Accordingly, a new alternative random matrix theory based approach to RCB is proposed in this work that explicitly addresses both the signature-mismatch problem and the limitations due to a finite sample-size. An improved performance is demonstrated via numerical simulations in the context of source power estimation in sensor array signal processing.