An Approach to Reduce the Overhead of Training Sequences in FDD Massive MIMO Downlink Systems

In frequency-division duplexing (FDD) massive multiple-input multiple-output (MIMO) systems, a large number of antennas are equipped at the base station (BS). In order to accurately estimate the downlink channel state information (CSI), a significant portion of time slots are usually invoked to send training sequences. To reduce the training time slots, we propose a dimension reduction method, which uses a matrix to divide and combine the transmit antennas into several groups. Noticed that the matrix is only related to the parameters of system and independent of CSI. As a beneficial result, a new equivalent system is created, where the training time slots can be decreased dramatically. Simulation results show that the new equivalent system can provide higher throughput than the original system because of the dimension reduction.