SBL-Turbo Compressed Sensing for Channel Estimation in OTFS Systems

‍ ‍In order to solve the problem of channel estimation accuracy degradation caused by Doppler shift in an orthogonal time frequency space （OTFS） system， this study investigated an SBL-Turbo compressed sensing channel estimation algorithm that jointly characterized the sparse nature of a wireless channel in a delay-Doppler domain. First， the sparse channel in the delay-Doppler domain was modeled as obeying a Gaussian prior model conditioned on the noise power. The mean and variance of the sparse channel were obtained by using a sparse Bayesian learning module， and an expectation-maximization algorithm was incorporated to update the parameters of the Gaussian prior model. Second， a linear minimum mean square error estimator module was introduced to improve the accuracy of the estimates， which re-estimated the posterior distribution of the sparse channel. The input values of the modules were decoupled from the output values by performing data processing on the channel posterior distribution estimated by each module， which in turn reduced the error propagation between blocks. Finally， the two modules used the Turbo structure to iteratively estimate the posteriori distribution of the channel and obtained the channel state information. Experimental results showed that compared with other estimation methods， this algorithm could improve the channel-estimation accuracy and BER performance of the system， and effectively solve the channel estimation problem caused by the Doppler shift in an OTFS system.