Downlink Training Overhead Reduction Technique for FDD Massive MIMO Systems

Abstract

In this letter, a novel minimum mean square error (MSE) based channel estimation framework is proposed to reduce the downlink channel training overhead in frequency division duplexing massive multiple-input multiple-output systems, where the overhead reduction is achieved through training only a subset of antennas and by exploiting the spatial correlation between the antennas at the base station. Closed-form expressions of the analytical MSE and the asymptotic MSE of the system are obtained. Furthermore, a perfect match between theoretical and simulation results is observed, where the channel training overhead can be reduced by half with an acceptable performance.