A novel self-adaptive density-based clustering algorithm for radar signal sorting

Radar signal sorting, which de-interleaves pulse sequences from different radar emitters, is one of the key techniques in Electronic Support Measures (ESM) system of modern warfare. As the electromagnetic environment in modern electronic wars is complex and dense, conventional multi-parameter de-interleaving techniques cannot achieve satisfactory accuracy. Although clustering algorithms have shown promising results in signal sorting, they all have their own weaknesses. In this paper, we propose a novel densitybased clustering algorithm targeting radar signal sorting problem. Our clustering algorithm determines the number of clusters and the density thresholds by the distribution of the data itself rather than depending on user experience as in K-Means or other existing density-based algorithms. Multiple density thresholds may be generated when necessary. We ran our algorithm on a set of simulation radar data. It outperforms DBSCAN-based signal sorting algorithm a lot in terms of accuracy, missing sorting rate and the error sorting in each set of parameters. In addition, it is observed that our algorithm is insensitive to the parameters, ratio factor, uniformity, and quantile. Thus, our proposed self-adaptive density-based clustering algorithm is promising in solving radar signal sorting problem in real electromagnetic environment.