Detection and Classification of GNSS Jammers Using Convolutional Neural Networks

Abstract

Global Navigation Satellite Systems (GNSSs) have been established as one of the most significant infrastructures in today's world and play an important role in many critical applications. It is known that the power of the GNSS signals at the receivers' antenna is extremely weak and the transmitted signals are vulnerable to interference, which can cause degraded positioning and timing accuracy or even a complete lack of position availability. Thus, it is essential for GNSS applications to detect interference and further recognize the types of it for the mitigation in GNSS receivers to guarantee reliable solutions. In this paper, the focus is on the automatic detection and classi-fication of chirp signals, known as one of the most common and disruptive interfering signals. The classifier is a Convolutional Neural Networks (CNN) based on multi-layer neural networks that operate on the representation of the signals in transformed domains, Wigner- Ville and Short Time Fourier transforms. The representation of signals is fed to a CNN algorithm to classify the different shapes of chirp signals. The proposed method is performed in two case-study scenarios: the monitoring and classification by a terrestrial interference monitor and from a Low-Earth-Orbit (LEO) satellite. The experimental results demonstrate that the CNN model has a classification accuracy of 93 % and can be a suitable approach to classify different shapes of chirp signals.