Open-Set Radar Emitter Recognition via Deep Metric Autoencoder

In the noncooperative electromagnetic environment, new radar emitters will emerge unexpectedly during the test phase, which brings the “open-set” radar emitter recognition (OS-RER). Conventional classifiers cannot identify new radar emitters that do not exist in the training data set. Therefore, in this article, a novel deep metric autoencoder (DMAE) is proposed for OS-RER. In DMAE, deep metric learning learns new nonlinear mappings in the metric space to measure the similarity between instances. The dual-path deep autoencoder is designed to reduce the open space risk by learning a low-dimensional manifold and a discriminative representation of known instances. Specifically, DMAE models known classes, and measures class belongingness through the reconstruction error of the AE and the entropy of the classifier. The deep metric network learns a more precise distance metric by minimizing the distance between the known class instances and the corresponding reconstruction. To accurately detect unknown instances, the classifier and the deep metric network are used together to preliminarily detect unknown instances. Finally, the detected unknown instances are used to further train the classifier to recognize the radar emitter in the open-set scenarios. The DMAE learns the discriminative representation through end-to-end learning. Extensive experiments conducted on real radar data sets and simulated radar data sets show that DMAE can identify unknown emitters and significantly outperforms existing open-set classification methods.

I. Introduction

Radar emitter recognition (RER) is an important subject in cognitive electronic warfare [1]. It can recognize the type of noncooperative radar emitters and the number of individuals for subsequent decision-making operations. As a typical pattern recognition task, RER can be well solved via machine learning algorithms [2]. Various types of models have been successfully used for RER, such as support vector machine [3], -nearest neighbor [4], random forest [5], etc. These methods typically extracted hand-crafted features from a database of known radar signals to train a classifier. In recent years, deep learning has been extensively applied to RER in an end-to-end manner [6], [7], [8], [9]. Due to its strong feature learning capabilities, it has proven to be a significant improvement over hand-crafted features and conventional classifiers, particularly in complex electromagnetic environments.