Beam Revisit Time Based Clustering Algorithm for Classification of Radars and Modes

Emitter identification techniques usually detect and identify different emitters by measuring primary features of the radar pulse data. The measured features are generally Pulse Amplitude, Pulse width, Pulse Frequency, Pulse Direction and Time of Arrival. The de-interleaved pulses are sorted on time of arrival (TOA) and one of the derived features Pulse Repetition Interval (PRI) is extracted using the traditional histogram-based methods by finding the cumulative differences and sequential differences. Emitter identification using supervised machine learning mechanisms is not feasible since labelling of pulse descriptor word (PDW) data is difficult. Also, deep learning-based Convolutional Neural Network (CNN) for radar identification and classification is available but the models are not effective for Active Electronically Scanned Antenna (AESA) based radars. In the modern complex electromagnetic environment radars receivers have the problem missing pulses and jittered parameters in frequency and PRI. Often emitters to avoid detection, use different counter techniques to vary the parameters and radiating patterns. This paper presents a novel technique to identify different emitters modes and characteristics using beam revisit times and illumination times as features by using two levels of clustering. This paper also brings out a unique technique of encoding the pulse data beams for identifying the types of modes of radar by finding the beam/burst patterns of the radar emitters.