Correlative coding approach for blind identification of MIMO systems

We address the problem of blind channel identification for multiple-input multiple-output (MIMO) finite impulse response (FIR) systems. This problem arises in the context of space division multiple access (SDMA) architectures for wireless communications. We develop a closed-form consistent estimator for the MIMO channel based only on second-order statistics. We rely on spectral asymmetry between the sources. This is obtained by correlative filters at each transmitter, which assign distinct spectral patterns to the random processes emitted by the users. We prove an identifiability theorem which guarantees the uniqueness of the channel estimate. The MIMO channel is identified by a second order moments matching strategy. We present an iterative algorithm which improves the closed-form estimate. Computer simulation supports our theoretical claims and illustrate the good performance of the proposed approach.