Enhancing Wireless Communication Systems through Robust RFI Detection and Classification

Abstract

Enhancing the Quality of Service (QoS) and security of wireless communication systems necessitates a robust Radio Frequency Interference (RFI) detection mechanism, enabling effective mitigation strategies. This study presents a novel multi-class Multi-Layer Perceptron (MLP) neural network designed for real-time classification of jamming signals in digital video broadcasting based on DVB-S2 standards. The communication signal is assumed to coexist with one of three major interference types: Continuous Wave Interference (CWI), Multiple CWI (MCWI), or Chirp Interference (CI). The proposed algorithm effectively distinguishes interference types and identifies the Signal of Interest (SoI). Principal Component Analysis (PCA) is employed to optimize feature selection, improving classifier performance. Additionally, various learning methods—online, full-batch, and mini-batch—are analysed to determine the most effective approach for real-time applications. A key contribution of this study is the creation of a novel real-time jamming signal dataset, distinct from those used in similar research, allowing for a more comprehensive evaluation of classifier robustness. Performance assessments, conducted at varying Signal-to-Noise Ratios (SNR), demonstrate the classifier's effectiveness in recognizing both known and unknown interference signals. Comparisons with the Support Vector Machine (SVM) technique reveal superior classification and detection capabilities of the proposed MLP-based design. These findings highlight the potential of the MLP approach for real-time RFI detection and mitigation, providing a significant advancement in ensuring the reliability and security of modern wireless communication systems.