A Jamming Recognition Algorithm Based on Deep Neural Network in Satellite Navigation System

Jamming recognition technology plays an important role in satellite navigation anti-jamming systems. Accurate and reliable recognition of jamming types is a necessary premise for adopting targeted anti-jamming means. To solve the problem of jamming recognition in satellite navigation systems, a jamming recognition algorithm based on deep neural network (DNN) is proposed in this paper. A set of jamming features with low complexity and high resolution is extracted in time, frequency and transform domains. Two jamming classifiers based on decision tree (DT) and DNN are constructed respectively, and their jamming recognition performance is compared. The simulation results indicate that when jamming-to-signal-and-noise ratio (JSNR) reaches 0 dB, the recognition rate of the DNN-based classifier for 12 types of typical suppressed jamming can reach more than 99%. Compared to the DT-based classifier, the DNN-based one improves the recognition rate of narrow band modulation, multi-tone, linear frequency modulation and sinusoidal frequency modulation jamming by more than 2 dB, better in recognition performance and easier to design. In the compound jamming situation, when JSNR reaches 10 dB, the recognition rate of the DNN-based classifier for all 10 types of compound jamming can reach more than 85%.