Closed-form blind channel estimation in orthogonally coded MIMO-OFDM systems: A simple strategy to resolve non-scalar ambiguities

We consider blind channel estimation in multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) systems with orthogonal space-time block code (OSTBC).We introduce a novel weighted covariance matrix of the received data which exploits all redundancies contained in the code. Based on the orthogonality properties of OSTBCs, we further show that a low-rank subspace mode applies to this matrix. Assuming a delay spread over the wireless channel smaller than the OFDM symbol duration, we prove that the corresponding signal subspace in time-domain collapses to a single principal eigenvector from which the channel taps can be uniquely determined. Unlike to previous methods, which suffer from ambiguities, for example, in the case of rotatable OSTBCs, our weighted covariance approach allows to resolve all non-scalar ambiguities provided that a specific transmitted symbol exhibits a higher power than the remaining symbols.