Estimation of Sparse Time Dispersive Channels in Pilot Aided OFDM Using Atomic Norm

We propose the use of the atomic norm minimization for the estimation of sparse time dispersive channels. The proposed estimation approach combines the atomic norm minimization, a super resolution method and the least squares (LS) method, and is intended for pilot aided channel estimation in OFDM systems. The combination of the atomic norm minimization and the super resolution method allows for gridless estimation of arbitrary delays of the individual paths in the channel impulse response, and the gains of those paths are then estimated using the LS method. To further improve the performance we also propose a version of the approach with reweighted atomic norm minimization. We compare the performance of the proposed approach with other grid based approaches for sparse channel estimation.