Radar Emitter Identification with Multi-View Adaptive Fusion Network (MAFN)

Radar emitter identification (REI) aims to extract the fingerprint of an and determine individual which it belongs. Although many methods have used deep neural networks (DNNs) for end-to-end REI, most them only focus on a single view signals, such as spectrogram, bi-spectrum, signal waveforms, so on. When electromagnetic environment varies, performance DNN will be significantly degraded. In this paper, multi-view adaptive fusion network (MAFN) is proposed by simultaneously exploring waveform ambiguity function (AF). First, original radar signals are separately feature extraction. Then, multi-scale feature-level module constructed features from waveforms AF, via Atrous Spatial Pyramid Pooling (ASPP) structure. Next, class probability modeled Dirichlet distribution perform decision-level evidence theory. Extensive experiments conducted two datasets, results show that MAFN can achieve accurate classification emitters more robust than its counterparts.