Signal detection for MIMO-ISI channels: a unitary linear recovery approach

In this paper, a new detector called unitary linear recovery (ULR) is proposed for MIMO-LSI channels with imprecise channel knowledge. The proposed method seeks the best match between the observation subspace and the Bezout signal subspace by optimizing over valid signaling matrices with structural, finite alphabet, and whiteness constraints. Forcing signal whiteness constraints result in a distinct distance measure (sum of singular values) to be used as the criterion. Algorithmically, the proposed method alternates between forcing signal orthogonality via searching for an optimal unitary linear recovery operator, and forcing Sylvester and finite alphabet constraints on the recovered outputs. Simulations show that ULR may outperform the MMSE detector with exact channel knowledge and ULR with successive interference cancellation gives a very attractive performance.