Distributed Compressed Sensing Aided Sparse Channel Estimation in FDD Massive MIMO System

Abstract

Massive multi-input multi-output (MIMO), which employs large number of antennas at the base station, can significantly boost the spectral efficiency and multiplexing gain. To fully exploit the huge array gain, the accurate channel state information is required at the transmitter side. However, the associated training overhead for downlink channel estimation consumes large amount of communication resource, especially for frequency division duplexing massive MIMO system. To address this issue, a distributed compressed sensing (DCS)-aided channel estimation approach is proposed, which fully exploits slow variation of the channel statistics in consecutive frames and spatially common sparsity within multiple subchannels in the frequency domain. Specifically, by exploiting the slow variation of the channel statistics, a hybrid training structure is proposed to probe the channel in the current frame based on the support information in previous frame. Then, a DCS-aided channel estimation algorithm, which combines least square method and DCS method, is proposed to estimate the two parts of channel vector in angular domain among different subcarriers. In addition, to effectively acquire the support information at the beginning of communication, a prior information estimation method is proposed by exploiting the uplink-downlink angular reciprocity. Simulation results demonstrate that the proposed approach outperforms the counterparts and is capable to significantly reduce the training overhead for channel estimation.