MFFNet: Multimodel Feature Fusion Networks for GNSS Interference Identification

The inherent vulnerabilities of Global Navigation Satellite Systems (GNSS) render them highly susceptible to various intentional or unintentional interferences. Accurate identification of these interferences is essential for selecting effective antijamming countermeasures. However, existing interference identification methods predominantly identifies interference based on the time-frequency images of interference signals. These methods suffers from significantly reduced identification performance when the time-frequency images of different interference signals exhibit high similarity. Therefore, this paper designs a Multimodal Feature Fusion Network (MFFNet) classifier, which identifies interference signals by fusing the time-frequency images, autocorrelation function images, and spectral flatness of the interference signals at the feature level. The tests demonstrate that even for low-power interferences within the range of -101 dBm to -110 dBm, the overall accuracy of the designed classifier exceeds 92% for nine common types of interference. This method can serve as an effective tool for GNSS interference identification.