This paper presents an efficient Semi-Blind (SB) approach for demixing multi-user mixtures, in a Multiple-Input Multiple-Output Orthogonal Frequency Division Multiplexing (MIMO-OFDM) communications system. A weighted hybrid cost function is introduced based upon the Multi-Modulus (MM) criterion, applied to unknown data, and the Least Squares (LS) criterion for pilot symbols. A separation procedure is put forward based on a block gradient descent approach, in which the step size is algebraically computed at each iteration and for each user. Besides a reduced computational complexity and an accelerated convergence, simulation results show that the proposed approach allows to mitigate the inherent ambiguity of fully-blind methods, and to withstand to the pilot contamination problem in massive MIMO systems.

Semi-blind source separation; Multi-modulus; MIMO-OFDM; Least squares; Least fourth