LAUSR

Accurate channel state information is of great importance in high-speed wireless communication. In this work, the problem of channel training for massive multiple-input multiple-output (MIMO) systems is investigated and a closed-loop channel training scheme is devised upon the maximum a posteriori (MAP) criterion. More specifically, at each channel use, the channel estimation is performed at the receiver according to the MAP criterion, and the expectation-maximization (EM) algorithm is employed to cope with the MAP estimation problem. Based on the parameter estimates, the predicted mean square error (PMSE) of the channel estimate is minimized to determine a preferred sounding beam for the next channel use, of which the index is fed back to the transmitter via a limited-rate channel. The performance superiority of the proposed approach is evidenced by comparing with various channel training schemes.