Application of Queuing Theory to the Estimation of the Feature Space for Radar Sources in Machine Learning

Abstract

The estimation of the feature space in analysis of radar signals (airplanes, ships, navigation stations, etc.), is an important element of machine learning. From the point of view of queuing theory, a mathematical model of a complex detected signal can be represented as the ordinary flow of events described by a Poisson distribution for randomly varying parameters of a signal. The paper demonstrates the orthogonality of the characteristic space of radar sources based on the entropy estimation. We show that the information measure of the discrepancy (Kullback-Leibler divergence) between intra-pulse modulation signals and frequency-modulated signals decreases with increasing pulse duration. Thus, in the recognition of radar sources using methods of machine learning, it is possible to limit ourselves to features without algorithms for processing intra-pulse modulation. Key words: radar source, queuing theory, Kullback-Leibler divergence, machine learning.