Return to Session B6 Next Abstract

Session B6: Receiver Design, Signal Processing, and Antenna Technology

Advancing Towards a Comprehensive GNSS Interference Management Framework: A Deep Learning Approach for Classification of FM Chirp Jammers
Yasmine Chaiben, University Gustave Eiffel; Syed Ali Kazim, IRT Railenium; Juliette Marais, University Gustave Eiffel
Location: Grand Ballroom GH
Date/Time: Thursday, May. 1, 4:05 p.m.

Global Navigation Satellite System (GNSS) plays a crucial role in modern applications, serving both civil and military needs by providing position navigation and timing services. However, GNSS signals are highly vulnerable to radio frequency interference (RFI) caused by jamming devices, which introduce signal distortions within the GNSS band. These disruptions pose significant challenges, leading to performance degradation or even complete denial of services. To suppress the effects of interference various countermeasures are developed, but their effectiveness varies depending on the type of interference encountered. Implementing an appropriate mitigation strategy requires a detection and classification module to identify different types of interference. This paper explores a deep learning-based approach for interference detection and classification using Convolutional Neural Networks (CNN). Two CNN models, VGGNet and ResNet, are implemented and analyzed for their effectiveness in accurately classifying the interference classes. Spectrogram images, which capture the signal’s time-frequency characteristics are used as network input to the network. The network is trained with 11 distinct classes of chirp-like signals. Our findings show that the trained models achieved an overall accuracy of 98%. In particular, certain classes such as Sawtooth, Triangular, Quadratic, Stepped FM, and Sinusoidal-shaped signals achieved 99.9% accuracy. Even when dealing with complex classes that can potentially overlap with others, such as the Gompertz the classification accuracy remained above 93%.
Index Terms—GNSS, Radio Frequency Interference, Convolutional Neural Networks, VGGNet, ResNet

Return to Session B6 Next Abstract