A comparison study of radar emitter identification based on signal transients

Abstract

Radar emitter identification has been studied for decades using library-based techniques that rely on pre-existing knowledge of parameters such as radio frequency (RF), pulse amplitude, pulse width, intentional pulse modulation type, or pulse repetition intervals. However, current radar emitter identification techniques will not be sufficient against cognitive radars due to their parameter agility and adaptability. In this study, five radar emitter identification fingerprints based on radar signal transients were analyzed and compared. These fingerprints include: (1) fractal dimension estimation of signal transients, (2) natural measures of signal transients, (3) polynomial regression of a signal transient energy trajectory acquired by its 4th order cumulants, (4) RF fingerprints based on the energy trajectory characteristics of signal transients, and (5) intrinsic shape of the rising edge of a pulse. The analysis and comparison were performed using K-Nearest Neighbours, Quadratic Discriminant Analysis, and relative entropy over a dataset from five different radar emitters. The advantages and drawbacks of each technique are highlighted. Our results show that (2), (4) and (5) achieve very competitive emitter identification performance using the selected radar datasets and classification algorithms. This study also demonstrates that the optimal emitter identification performance is dependent on the combination of RF fingerprints and classification algorithms.