Jamming signals classification using convolutional neural network

Abstract

In the complex electromagnetic environment, satellite communication links will suffer kinds of interference and jamming, including deception jamming, suppression jamming, and communication network interference. Each of these can be subdivided into a more accurate signal. For example, suppressed jamming includes audio jamming, narrowband jamming and sweep jamming and so on. It's necessary to detect and classify the jamming and interference in the communication link. This paper proposes an automatic jamming signal classification method using a convolutional neural network (CNN). We use five types of jamming mode as input signals including audio jamming, narrowband jamming, pulse jamming, sweep jamming and spread spectrum jamming. Considering the characteristic of CNN, after verifying the feasibility of our method, it's easy to extend CNN training set and apply to more signals. The feature automatically extracted by CNN has a strong robustness against a large range of jamming noise rate (JNR). Single jamming classification and coexist jamming classification simulation results show that the classification accuracy of CNN is remarkable.