Small-Sample-Support Channel Estimation For Massive Mimo Systems

We consider the problem of blind channel estimation with minimal pilot signaling in multi-cell multi-user MIMO systems with very large antenna arrays at the base station. We develop a least-squares (LS)-type algorithm that iteratively extracts channel and data estimates in short-data record multicell massive MIMO environments with no prior channel state information. The proposed algorithm utilizes a novel initialization step that is based on auxiliary-vector (AV) subspace decomposition. Simulation studies show that for pilot signaling of about 4%, information data extraction can be achieved with lower probability of error than eigendecomposition-based initialization techniques, while for observation records of sufficient length it nearly attains the error rate performance achieved with complete knowledge of the channels.