A Cross-Domain Method for OTFS Channel Estimation and Equalization

To detect Orthogonal Time Frequency Space (OTFS) modulated data in high-speed scenarios, effective channel estimation is of great significance for successful transmission. In this letter, not limited to a single domain, we propose a cross-domain method, obtaining reliable channel estimation with much lower computational complexity by applying channel estimation in the Doppler-delay (DD) domain and equalization in the time-frequency (TF) domain. We propose a threshold estimation method based on noise intensity and Cyclic-Prefix length. To handle mismatch of pulses and TF grid caused by one-tap equalizer, we estimate the self interference using pilot and guard symbols. We design a sliding window structure for deep learning to accelerate the model convergence and improve the performance. Simulating results validate the effectiveness of the proposed solution under different indicators.