Towards a Faster GNSS Interference Classification: a GRU-Based Approach using Spectrograms

Global Navigation Satellite Systems (GNSS) are critical for Positioning, Navigation, and Timing (PNT) services, but their signals are highly vulnerable to intentional jamming. Traditional interference detection methods, such as spectral analysis and statistical thresholding, struggle to adapt to dynamic interference conditions. Recent advances in Artificial Intelligence (AI) using deep learning techniques, have enabled more robust classification approaches. This study proposes a Gated Recurrent Unit (GRU)-based framework for GNSS interference classification, offering a more lightweight alternative to convolutional neural networks (CNNs), which rely on spectrogram images as inputs. The GRU model directly processes sequential spectrograms, enabling efficient real-time classification with a reduced computational overhead. Experimental validation using datasets collected in a controlled environment, along with an implementation on a low-cost embedded device, demonstrated high classification accuracy and improved efficiency over CNN-based approaches. The findings highlight the potential of AI-driven methods for scalable and real-time GNSS interference monitoring, particularly in resource-constrained environments.