Constrained ML algorithms for semi-blind MIMO channel estimation

We propose and study algorithms for constrained maximum-likelihood (ML) estimation of a unitary matrix in the context of semi-blind multi-input multi-output (MIMO) channel estimation. The flat-fading r×t MIMO channel matrix, H, for r≥t can be decomposed as the matrix product H = WQ^H, where W is a whitening matrix and Q is a unitary rotation matrix. Exclusive estimation of Q from pilot symbols has been shown potentially to achieve a 3 dB or greater improvement in terms of channel estimation accuracy. We develop and present the OPML, IGML and ROML algorithms for the constrained estimation of the unitary matrix Q; they are appropriate for a variety of scenarios, e.g., orthogonal pilots, low complexity, etc. Simulation results are provided to demonstrate the efficacy of the algorithms.