Channel Estimation and Localization for mmWave Systems: A Sparse Bayesian Learning Approach

Millimeter wave (mmWave) systems have potential advantages for high accuracy localization due to their large bandwidth and highly directional and sparse channel. The existing schemes on mmWave system localization only explore the channel's angular sparsity and restrict the representation of angular parameters to fixed grids. In this paper, we propose a joint sparse channel model with adjustable grids of angles and delays of mmWave paths, which not only enables exploiting the joint sparsity in angular and delay domain, but also mitigates the quantization error due to fixed grids. We formulate the estimation of channel, angles and delays as a sparse Bayesian learning problem and propose to solve it using an in-exact block minorization-maximization algorithm and then determine the user location. We verify the performance of the algorithm by simulations.