A One-Class Bayesian Algorithm for Radio Frequency Interference Detection in Microwave Radiometry

Measurements of Earth-observing radiometers have been reported to be contaminated by radio frequency interference (RFI) due to active emissions in the densely occupied radio spectrum. This study presents a multi-dimensional, one-class Bayesian algorithm to detect and eliminate such RFI. The proposed algorithm was trained with RFI-free measurements only and operated in a feature space where the separation between RFI-free and RFI-contaminated measurements was maximized. Using standard metrics, the performance of the Bayesian algorithm was evaluated, and it was demonstrated that the one-class detection algorithm presented in this paper can outperform the existing state-of-the-art RFI detection algorithms, specifically for the low interference to noise ratio (INR) cases.