Answers are grounded in this paper and cite their sources.
Extensions of MVDR and Capon estimation techniques are presented for the situation in which the signal is either rank-one of unknown orientation in a subspace or multi-rank. Only signal-plus-noise snapshots are available. The relationships among linearly and quadratically-constrained approaches are clarified and a unified treatment is given that includes both direct and sidelobe canceller architectures. The unifying component is the multi-rank MVDR beamformer followed by post processing. Detection statistics are presented for the situation in which there is no signal-free training data. Simulations are used to compare rank-one and multi-rank performance. I.
No takes yet. Share an insight, caveat, or question.
Cox et al. (2006) studied this question.