Frequency Hopping Signal Sorting Based on Spectrum Monitoring Data by Adaptive DBSCAN

In non-cooperative scenarios, mining spectrum monitoring data to sort electromagnetic signals is significant for communication behavior detection, network topology inference, communication countermeasure, etc. To overcome the obstacle of signal reconnaissance from frequency hopping, this paper proposes a signal sorting algorithm based on spectrum monitoring data by clustering physical layer features with adaptive DBSCAN. Firstly, we analyze frequency hopping communication properties and spectrum monitoring operational mode. Then, we mine and extract physical layer features of frequency hopping signal, without prior knowledge about networks. Next, in order to accomplish signal sorting, the DBSCAN algorithm which is suitable for streaming data is studied. For the problem that manual selection of global parameters is required, DBSCAN is parametric adaptively improved, whose global parameters are determined based on the gradient matrix of dataset KNN distribution. Then, the frequency hopping signal sorting algorithm is proposed on the basis of feature extraction and DBSCAN improvement, on which simulations of signal sorting are carried out. Finally, extensive simulations under different signal sorting algorithms are conducted to validate the effectiveness of the proposed algorithm. It shows that the proposed algorithm has excellent performance on frequency hopping signal sorting.