A Model-Based Neural Network for PRACH Joint Time-Frequency Synchronization
Hamidreza Khaleghi, Stéphane Paquelet · 2023
In this paper, we address the limitations of existing approaches for time-frequency synchronization in the wireless communications. We propose a novel parameter estimator that achieves the Cramer-Rao Lower Bound (CRLB) performance, thereby overcoming the inherent limitations of traditional methods. Our approach entails the development of an accurate analytical model of the PRACH signal, followed by the creation of a Maximum Likelihood Estimator (MLE). One of the key innovations of this work lies in the implementation of the estimator on a real-time platform using a model-driven Neural Network. By leveraging the power of neural networks, we demonstrate an innovative approach that enhances the time and frequency offset estimation and consequently the synchronization accuracy. Our extensive evaluations and exhaustive tests reveal promising results, validating the effectiveness of our proposed approach.