Algorithm for Radar Signal Pre-sorting Based on Correlation Distance Differential
Shuo Zhou, Zhiyu Qu, Xujie Li · 2023
Radar signal pre-sorting plays a crucial role in the entire radar signal sorting process. Currently, the primary approach to radar pre-sorting is unsupervised clustering. However, common clustering algorithms face challenges: they may struggle to handle "non-cluster" shaped data caused by parameter variations; exhibit high algorithmic complexity, leading to time-consuming processes that fail to meet real-time requirements; necessitate pre-configured parameters, wherein improper settings can result in clustering errors; or encounter difficulties in dealing with the constant influx of new radar styles, exhibiting limitations in various aspects. To address these issues, this paper proposes a radar pre-sorting algorithm based on Correlation Distance Differences(CDD). This algorithm computes the correlation between pulses, representing it as correlation distance, performs a first-order differentiation on the sorted distances, and subsequently identifies local extreme points. These points indicate boundaries between different classes of radars. Notably, the algorithm doesn't require preset parameters, boasts a straightforward principle, and maintains a relatively low computational complexity. Particularly, it excels in clustering performance for special radar styles characterized by changing carrier frequencies and complex pulse repetition patterns. Simulation experiments demonstrate the algorithm's robust clustering performance and high accuracy, achieving up to over 90% accuracy in ideal noise-free simulation environments. As a result, the algorithm holds promise for practical engineering applications.