A Two-Stage OMP-DL Approach with Multi-Scale GRU for OTFS Channel Estimation in V2X

Orthogonal Time Frequency Space (OTFS) modulation is an emerging multicarrier technique. It multiplexes and transmits signals in the Delay-Doppler (DD) domain, offering a robust solution for Vehicle-to-Everything (V2X) communication scenarios. Accurate channel estimation is critical for achieving high-performance OTFS communication within these environments. Traditional channel estimation methods often result in high Peak-to-Average Power Ratio (PAPR). They also tend to overlook the nonlinear effects of High-Power Amplifiers (HPAs) on the system. To address these challenges, this paper proposes a two-stage OMP-DL approach combining traditional algorithms with deep learning. Our proposed method first employs the traditional Orthogonal Matching Pursuit (OMP) algorithm for initial estimation of preambles and pilot signals. Next, a Deep Neural Network (DNN) standardizes these initial results. This standardized data then feeds into a Multi-Scale Gated Recurrent Unit (Multi-Scale GRU). The Multi-Scale GRU extracts features from its small-, medium-, and large-scale components for subsequent feature fusion. Finally, a denoising module produces the optimized estimation result. Simulation results demonstrate our algorithm's effectiveness. It achieves lower Bit Error Rate (BER) and Normalized Mean Square Error (NMSE). Furthermore, it significantly mitigates the problem of high PAPR.