Deep Neural Networks for Radar Waveform Classification

TL;DR

This paper develops deep neural networks capable of classifying radar waveforms from raw I/Q data, effectively handling noise, synchronization issues, and superimposed pulses, significantly outperforming previous methods.

Contribution

The paper introduces robust DNN architectures for radar waveform classification that handle noise, synchronization, and superimposition, achieving over 100x error-rate reduction.

Findings

Over 100x reduction in error-rate compared to previous methods

Effective classification of superimposed radar pulses

Robust performance across varying SNR and pulse widths

Abstract

We consider the problem of classifying radar pulses given raw I/Q waveforms in the presence of noise and absence of synchronization. We also consider the problem of classifying multiple superimposed radar pulses. For both, we design deep neural networks (DNNs) that are robust to synchronization, pulse width, and SNR. Our designs yield more than 100x reduction in error-rate over the previous state-of-the-art.