Iterative Receiver Design With Off-the-Grid Sparse Channel Estimation

In this work we design an iterative receiver that incorporate sparse channel estimation. State-of-the-art sparse channel estimators simplify the estimation problem to be a finite basis selection problem by restricting the multipath delays to a grid. Our main contribution is a receiver that is released from such a restriction; the delays are off-the-grid, i.e., they are estimated and tracked directly as continuous values. As a result our receiver does not suffer from the leakage effect, which destroys sparsity when the delays are restricted to a grid. We use the unifying framework of combined belief-propagation and mean-field. All parameters in the receiver are inherently estimated. The receiver outperforms iterative receivers embedding state-of-the-art sparse channel estimators in terms of both mean- squared error of the channel estimate and bit error rate. We also demonstrate that our receiver design allows for a significant reduction in the number of pilot signals, without incurring any increase in bit error rate. The receiver also adapts well to situations where the sparse channel assumption is violated; in this case its bit error rate is comparable to that of an iterative receiver that uses minimum mean-squared error channel estimation.