Ambiguity Function Based Radar Waveform Classification and Unsupervised Adaptation Using Deep CNN Models

Abstract

We present a robust generalized approach to phase and frequency modulated LPI Radar waveform classification and adaptation, inspired by deep convolutional neural architectures. We use a complex Ambiguity Function matrix as a pre-processing step, following which, a waveform classification, or adaptation to unlabeled reference target domains, is performed. We test our method on a wide range of tasks, datasets, and different signal distributions. Our method surpasses the state-of-the-art performance on classification problems on multi-encoding, multi-feature datasets, in diverse and challenging conditions. Our novel approach to an unlabeled Radar waveform adaptation reveals impressive classification improvements to domain shifted unlabeled signals.