Blind identification of MIMO systems based on source correlative filtering

We introduce a closed-form blind channel estimator for multiple-input multiple-ouput (MIMO) finite impulse-response (FIR) systems, based only on second-order statistics (SOS). We rely on correlative filters at each transmitter to induce a spectral asymmetry between the users. No additional power or bandwidth, nor synchronization between the sources, are required, and the original data rate is maintained. We show that, under a simple spectral condition on the transmitted random processes, this data preprocessing makes the MIMO channel uniquely determined (up to a phase offset per user) from the SOS of the MIMO system outputs. The closed-form algorithm which attains this channel identifiability bound is discussed.