RFI mitigation technique using statistical models and deep learning on FPGA and GPU

Abstract

This study tackles the enduring problem of radio frequency interference (RFI) in vital fields like satellite systems, wireless communications, and radio astronomy. We suggest a novel hybrid method that combines a Convolutional Neural Network’s (CNN) capacity for pattern identification with the statistical stability of the Median Absolute Deviation (MAD). The goal of combining MAD and CNN is to offer a flexible and adaptable RFI mitigation system that can identify and reduce interference in a variety of situations. The study progresses by thoroughly examining how RFI affects various industries, highlighting the necessity of effective mitigation techniques. The hybrid approach is intended to reduce interference in wireless communications and satellite systems, where uninterrupted signal delivery is crucial, as well as in radio astronomy, where precision and clarity are crucial. The iterative use of MAD, which introduces a paradigm shift in interference mitiga- tion, is a crucial component of this work. Examples of Wideband RFI cases show how applying MAD repeatedly can provide iterative refinement in the suppression of interfer- ence. Together with quantitative measurements like signal-to-noise ratio (SNR) and mean square error (MSE) to evaluate signal fidelity, the study also provides visual evidence of the hybrid approach’s impact.