GNSS Interference Signal Recognition Based on Deep Learning and Fusion Time-Frequency Features

Abstract

Global navigation satellite system (GNSS) can provide time, location, speed information at anytime and anywhere. Moreover, it can realize navigation, high-precision positioning, and time service in real time. Because satellite navigation signals are susceptible to various types of intentional or unintentional interference, which has great influence on navigation application security, it is urgent to develop GNSS interference monitoring system to provide decision basis for responding to interference. Interference signal modulation recognition in GNSS is helpful to identify users and distinguish sources of interference, and increase situational awareness. In order to improve the recognition accuracy in the case of low interference-signal power rate (ISR), a new recognition method based on fusion time-frequency features and convolutional neural network is proposed. The new time-frequency transform combining short-time Fourier transform (STFT) and WignerVille distribution (WVD) was proposed aiming at the problem of mutual restriction between time-frequency focusing and cross term suppression of single time-frequency transform. The new fusion time-frequency features combine the characteristics of STFT without cross term and WVD with high time-frequency resolution. Taking time-frequency image as recognition feature provides a new method for GNSS interference signal recognition. By learning the time-frequency characteristics of interference signal, the modulation recognition problem is transformed into image recognition problem. Convolutional neural networks (CNNs) are widely used in the field of computer vision, it shows excellent performance in image classification and image recognition. This article uses convolutional neural network to complete the classification of interference signals. Through pre-process, fusion time-frequency features are converted into dimensions suitable for CNN training and learning, and then use the CNN classifier in this paper to complete modulation recognition. The experimental results show that the new combined time-frequency transform features can realize the classification and identification of typical modulated interference signals, and compared with the single time-frequency transform, the recognition accuracy of interference modulation is improved, and the recognition effect is better.