Recursive MMSE estimation of wireless channels based on training data and structured correlation learning

We present a novel block-recursive technique for training based channel estimation that augments MMSE channel estimation with structured correlation learning to obtain the required statistics. We provide formulations for both stationary and nonstationary wireless channels along with a robust initialization procedure. The method is efficient, applicable in a variety of communications schemes (MIMO, multiuser, OFDM), and has performance superior to that of RLS from which it differs in an important way. Simulation examples show nearly optimum performance with particular gains for strongly correlated channels