EM-based estimation of doubly selective channel with a basis expansion model

An expectation-maximization (EM)-based estimator for doubly selective channel parameterized by basis expansion model (BEM) is proposed, which converts the problem of estimating a multipath channel to parallel channel estimations. The convergence property of the algorithm has been carefully investigated, which shows that global convergence is always guaranteed. A novel initialization strategy is proposed to accelerate convergence speed and reduce computational cost, the effectiveness of which is demonstrated through simulations.