Wireless Jamming Identification via Hypergraph Partitioning and Spectral Clustering

Wireless jamming identification (WJI) is crucial for enhancing the robustness of radio systems in the complicated electromagnetic environments. In this paper, we present an WJI approach that leverages the feature extraction and the hypergraph learning to improve identification performance. Specifically, the spectrum features of the received signal are first extracted to represent the essential distinctions of the jamming signals. Then, we construct the hypergraph model utilizing the extracted feature vector to capture the multi-way relationships within the signals, followed by the hypergraph partitioning and the spectral clustering to determine the unknown jamming type. Numerical results show that the proposed hypergraph learning identification approach outperforms the baseline methods, achieving high identification accuracy with well-designed features and improving the robustness to the noise.