Fractional Delay-Doppler Channel Estimation for OTFS Systems Based on Segmentation Technique and Sparse Bayesian Learning

Abstract

In this letter, a doubly fractional channel estimation (CE) scheme based on segmentation and sparse Bayesian learning (SBL) is proposed for orthogonal time-frequency space systems. The scheme introduces delay-Doppler segmentation factors adaptively selected based on channel quality to construct an extended measurement matrix, complemented by a complexity reduction method that makes use of the characteristics of diagonal matrices. Additionally, the received signal is decomposed into two branches along different dimensions, and an SBL combination method with weighted average fusion and a fixed-point update rule is employed to further improve the performance. Simulation results show that the proposed scheme achieves enhanced CE performance and lower complexity, compared with the baseline schemes.