Classification of airborne radar signals based on pulse feature estimation using time-frequency analysis

Abstract

This paper describes the realization of an airborne radar signal type analysis and classification (ARTAC) system that uses spectrograms to obtain time-frequency representation ((t,f) representation) and then apply the related analysis tools, such as the instantaneous energy and frequency, and time-frequency marginal, to estimate the various signal characteristics. The estimated parameters are used as input to a rule-based classifier that classifies the signal appropriately. Monte-Carlo simulation is then conducted to quantify the accuracy of signal classification at various signal-to-noise ratios (SNRs) in additive white Gaussian noise (AWGN). The methodology used achieves 90% classification accuracy at SNR of 6 dB irrespective of the identity of the signal. The performance and computational complexity (CC) of the system are also addressed in an electronic support (ES) operating scenario.