Jammer Classification in GNSS Bands Via Machine Learning Algorithms

This paper proposes to treat the jammer classification problem in the Global Navigation Satellite System bands as a black-and-white image classification problem, based on a time-frequency analysis and image mapping of a jammed signal. The paper also proposes to apply machine learning approaches in order to sort the received signal into six classes, namely five classes when the jammer is present with different jammer types and one class where the jammer is absent. The algorithms based on support vector machines show up to 94 . 90 % accuracy in classification, and the algorithms based on convolutional neural networks show up to 91 . 36 % accuracy in classification. The training and test databases generated for these tests are also provided in open access.