Automatic Classification of Frequency-Modulated Radar Waveforms Under Multipath Conditions

Abstract

Techniques for automatic modulation classification (AMC) of radar signals are crucial for spectrum sensing, passive bistatic radars, and electronic intelligence (ELINT). Most of the existing AMC algorithms were evaluated using solely synthetic data. Meanwhile, signals intercepted in real-life scenarios are extra distorted due to the multipath propagation. This article presents the novel, pattern recognition AMC method for frequency-modulated radar signals, with improved resistance to noise and multipath. This has been achieved by a multistage feature selection process, involving over a dozen of popular radar signal metrics. Based on the feature selection results, the two advanced waveform features were utilized in the designed algorithm, i.e., quasi-maximum likelihood (QML) instantaneous frequency (IF) estimate and fractional Fourier transform (FrFT) profile. The proposed AMC method has been evaluated using both synthetic and real data. The obtained results show that proposed classification framework achieves an overall accuracy of 93.6% for a set of real-life signals acquisitions, corrupted both by noise and multipath influence.