A Novel Deep Learning-Based Coarse-to-Fine Frame Synchronization Method for OTFS Systems

TL;DR

This work addresses frame synchronization for OTFS in high-mobility channels by formulating STO estimation as a DL-based classification task and proposing a non-data-aided coarse-to-fine ResNet architecture. The method halves the search space through a two-stage process: first identifying a coarse segment $\,\hat{\theta}_t$ among $N$ segments of length $M$, then refining within the segment to obtain $\,\hat{\theta}_d$, with the final STO given by $\hat{\theta} = \hat{\theta}_t M + \hat{\theta}_d$. A custom 1D ResNet with multi-scale kernels and residual connections extracts temporal features from OTFS pilot periodicity in the DT domain, trained with AdamW over 500 epochs. Extensive simulations across AWGN, Rayleigh, and EVA channels show superior accuracy and lower RMSE compared with conventional preamble-based and 2D autocorrelation methods, particularly at low SNRs, while reducing computational complexity relative to a one-stage DL classifier. The results demonstrate a practical, low-overhead mechanism for OTFS frame synchronization with potential for real-time deployment on resource-constrained devices using GPU acceleration.

Abstract

Orthogonal time frequency space (OTFS) modulation is a robust candidate waveform for future wireless systems, particularly in high-mobility scenarios, as it effectively mitigates the impact of rapidly time-varying channels by mapping symbols in the delay-Doppler (DD) domain. However, accurate frame synchronization in OTFS systems remains a challenge due to the performance limitations of conventional algorithms. To address this, we propose a low-complexity synchronization method based on a coarse-to-fine deep residual network (ResNet) architecture. Unlike traditional approaches relying on high-overhead preamble structures, our method exploits the intrinsic periodic features of OTFS pilots in the delay-time (DT) domain to formulate synchronization as a hierarchical classification problem. Specifically, the proposed architecture employs a two-stage strategy to first narrow the search space and then pinpoint the precise symbol timing offset (STO), thereby significantly reducing computational complexity while maintaining high estimation accuracy. We construct a comprehensive simulation dataset incorporating diverse channel models and randomized STO to validate the method. Extensive simulation results demonstrate that the proposed method achieves robust signal start detection and superior accuracy compared to conventional benchmarks, particularly in low signal-to-noise ratio (SNR) regimes and high-mobility scenarios.

Figures (9)

• Figure 1: OTFS system block diagram.
• Figure 2: DD domain signal with embedded pilot and its counterpart in the DT domain.
• Figure 3: The real part of the signal in DT grid. (a) Transmitted signal. (b) Received signal.
• Figure 4: Time domain signal sampling.
• Figure 5: Coarse-to-fine frame synchronization method.