2D Structured Turbo Compressed Sensing for Channel Estimation in OTFS Systems

High-speed mobility is a challenging scenario in wireless communications since it involves fast-varying channels, in which orthogonal frequency-division multiplexing (OFDM) could degrade significantly without accurate Doppler compensation. Recently, a modulation called orthogonal time frequency space (OTFS) was proposed for the specific scenario. The advantages are two-fold, i.e., the orthogonality in the time-frequency domain and the clustered sparsity of channel information in the delay-Doppler domain. In this paper, we propose a two-dimensional structured Turbo compressed sensing (CS) algorithm based on our previous work on Turbo CS for fast time-varying channel estimation. Markov random field is applied to model the clustered sparse prior of the channel in the two-dimensional delay-Doppler domain. In contrast to the conventional Turbo CS, a two-dimensional structured minimum mean-square error (MMSE) estimator is designed by message passing over a Markov random field. The convergence and the MSE performance of the proposed algorithm is evaluated, and the simulation results validate that the proposed algorithm outperforms the existing algorithms.