OTFS Channel Estimation via 2D Off-grid Decomposition and SBL Combination

Abstract

Fractional channel parameter is one of the main factors deteriorates the performance of channel estimation in orthogonal time-frequency space (OTFS) system. Several works had explored the one dimensional (1D) and two dimensional (2D) off grid methods to handle the fractional delay and Doppler shift. However, the 1D off-grid method is computationally complex and the 2D off-grid method suffers performance degradation. One reason of performance degradation in 2D off-grid method is the tandem off-grid distortion introduced during the estimation of intermediate channel matrix in its first step. To mitigate the tandem off-grid distortion, a 2D off-grid decomposition and SBL combination method is proposed in this paper. In our method, the received symbols are firstly processed by two paralleling branches, i.e., the U-branch and the V-branch, based on the decomposition of the OTFS symbols in different dimension. The two branches are used to estimate the off-grid parameters and the channel matrix simultaneously and independently, leading to complementarity in representing the received symbols. Then the obtained two channel matrices and the off-grid parameters are combined to iteratively construct the final channel matrix based on the sparse Bayesian learning (SBL) framework. Through the off-grid decomposition and SBL combination, the channel matrix is better represented and higher accuracy could be achieved. The simulation results show the effectiveness of the method.