An Improved Adaptive Radar Signal Sorting Algorithm Based on DBSCAN by a Novel CVI

Abstract

Radar signal sorting is a crucial step in radar signal processing, and its accuracy directly affects the progress of electronic warfare. Numerous machine learning algorithms, such as K-Means and density-based spatial clustering of applications with noise (DBSCAN), have been applied to radar signal sorting. However, these algorithms were not enough suitable for radar pulse data with overlapping in space and required prior information about the number of radar signal sources. This paper proposed an adaptive radar signal sorting algorithm based on DBSCAN. Inspired by the density peak clustering algorithm's idea, we integrated it into Davies-Bouldin index (DBI) and then proposed a novel cluster validity index (CVI) density peaks Davies-Bouldin index (DPDBI) at the same time. The algorithm used the kernel density estimation method to determine the DBSCAN algorithm parameter range and obtained the optimal parameters to complete clustering based on the DPDBI calculation results. The algorithm could determine the clustering cluster number and complete clustering without any parameter input or overlapping signal parameters. Experiments were conducted on radar signal datasets, real datasets, and synthetic datasets to demonstrate the effectiveness of the proposed cluster validity index DPDBI. The clustering results obtained from experiments on radar signal datasets have an average accuracy of over 95%, which was higher than the current existing algorithms, demonstrating the effectiveness and superiority of the new algorithm DPDBI-DBSCAN.