Improved Label Design for Timing Synchronization in OFDM Systems against Multi-path Uncertainty

TL;DR

This paper introduces a learning-based timing synchronization method for OFDM systems that uses a novel training label design to improve accuracy under multi-path uncertainty, leveraging prior LOS information.

Contribution

It proposes a new training label design that highlights the ISI-free region, improving the robustness of machine learning-based timing synchronization in multi-path environments.

Findings

Enhanced synchronization accuracy under multi-path conditions.

Effective use of LOS information for label correctness.

Numerical results show improved performance over traditional methods.

Abstract

Timing synchronization (TS) is vital for orthogonal frequency division multiplexing (OFDM) systems, which makes the discrete Fourier transform (DFT) window start at the inter-symbol-interference (ISI)-free region. However, the multi-path uncertainty in wireless communication scenarios degrades the TS correctness. To alleviate this degradation, we propose a learning-based TS method enhanced by improving the design of training label. In the proposed method, the classic cross-correlator extracts the initial TS feature for benefiting the following machine learning. Wherein, the network architecture unfolds one classic cross-correlation process. Against the multi-path uncertainty, a novel training label is designed by representing the ISI-free region and especially highlighting its approximate midpoint. Therein, the closer to the region boundary of ISI-free the smaller label values are set, expecting to locate the maximum network output in ISI-free region with a high probability. Then, to guarantee the correctness of labeling, we exploit the priori information of line-of-sight (LOS) to form a LOS-aided labeling. Numerical results confirm that, the proposed training label effectively enhances the correctness of the proposed TS learner against the multi-path uncertainty.