Dictionary Learning Based Sparse Channel Representation and Estimation for FDD Massive MIMO Systems

This paper addresses the problem of uplink and downlink channel estimation in FDD Massive MIMO systems. By utilizing sparse recovery and compressive sensing algorithms, we are able to improve the accuracy of the uplink/downlink channel estimation and reduce the number of uplink/downlink pilot symbols. Such successful channel estimation builds upon the assumption that the channel can be sparsely represented under some basis/dictionary. Previous works model the channel using some predefined basis/dictionary, while in this work, we present a dictionary learning based channel model such that a dictionary is learned from comprehensively collected channel measurements. The learned dictionary adapts specifically to the cell characteristics and promotes a more efficient and robust channel representation, which in turn improves the performance of the channel estimation. Furthermore, we extend the dictionary learning based channel model into a joint uplink/downlink learning framework by observing the reciprocity of the AOA/AOD between the uplink/downlink transmission, and propose a joint channel estimation algorithm that combines the uplink and downlink received training signals to obtain a more accurate channel estimate. In other words, the downlink training overhead, which is a bottleneck in FDD Massive MIMO system, can be reduced by utilizing the information from simpler uplink training.