LAUSR

In this letter, we reveal that in the massive multiple-input multiple-output system with large bandwidth, sub-channels of orthogonal frequency division multiplexing share approximately sparse common support due to the frequency difference of subcarriers. We use the approximate message passing with nearest neighbor sparsity pattern learning (AMP-NNSPL) algorithm to adaptively learn the underlying structure for improving the accuracy of channel estimation, where the learning strategy is newly derived by solving an optimization problem. In addition, the performance of the AMP-NNSPL is well predicted by the state evolution. Simulation results demonstrate the superiority of the algorithm in systems with large bandwidth.