Off-Grid Channel Estimation With Sparse Bayesian Learning for OTFS Systems

This paper proposes an off-grid channel estimation scheme for orthogonal time-frequency space (OTFS) systems adopting the sparse Bayesian learning (SBL) framework. To avoid spreading caused by fractional delay and Doppler shifts to fully exploit sparsity in delay-Doppler (DD) domain, we estimate original DD domain response rather than effective as commonly adopted literature. OTFS is firstly formulated a one-dimensional (1D) signal recovery (SSR) problem based on virtual sampling grid defined space, where on-grid components of are separated estimation. In particular, jointly determined entry indices with significant values recovered vector. Then, corresponding modeled hyper-parameters proposed SBL framework, which can be estimated via expectation-maximization method. strike balance between performance computational complexity, further propose two-dimensional (2D) SSR decoupling shift estimations. our developed 1D 2D SBL-based algorithms, updated alternatively computing conditional posterior distribution channels, exploited reconstruct channel. Compared method, method enjoys much lower complexity while only suffers slight degradation. Simulation results verify superior schemes over state-of-the-art schemes.