Mitigating Sampling Frequency Offset in OFDM: A Comparative and Machine Learning-Based Approach

Document Type : Original Research Papers

Authors

1 EE Dept. Benha faculty of engineering, Benha University, Benha, Egypt.

Abstract

This paper addresses the critical issue of Sampling Frequency Offset (SFO) in Orthogonal Frequency Division Multiplexing (OFDM) systems, which arises from mismatches between the sampling rates of the transmitter and receiver. Such discrepancies disrupt subcarrier orthogonality, leading to significant performance degradation, including Inter-Carrier Interference (ICI), phase distortion, and increased Bit Error Rate (BER). To ensure reliable data transmission, accurate SFO estimation and compensation are essential. The study examines four widely used SFO estimation techniques: the Phase Difference (PD) method, the Correlation-Based (CB) method, the Phase Difference Weighted by Subcarrier Index (PD-WSI) method, and the Hybrid Estimation (H-EST) method. Additionally, it introduces novel machine learning-based approaches—Linear Discriminant Analysis (LDA)-based, Kernel Support Vector Machine (KSVM)-based, and Artificial Neural Network (ANN)-based SFO estimators—designed to enhance synchronization accuracy. Comparative evaluations demonstrate that these proposed methods significantly outperform conventional and hybrid techniques by achieving lower Root Mean Square Error (RMSE), thereby effectively mitigating SFO-induced impairments and improving overall OFDM system performance.

Keywords

Main Subjects

Benha Journal of Applied Sciences
Volume 10, Issue 4 - Serial Number 4
Engineering Sciences
April 2025
Pages 131-150

Files

Share

How to cite

Statistics