A Parallel Neural Network-based Scheme for Radar Emitter Recognition

Passive radar systems are used in the military for intelligence gathering, threat detection and as a support to electronic attack systems. Therefore, radar emitter recognition is a crucial task of reconnaissance systems for accurately identification of hostile threats. However, this problem is challenging due to the complicated noisy electromagnetic environment as well as the increasing complexity of modern radar signals. In this paper, we introduce a novel deep neural network-based scheme, named ParallelNet for the recognition of different radar types. In our approach, I/Q samples and radar pulse features extracted from received wideband signal are inputs of two parallel subneural networks. The outputs of sub-networks are subsequently combined to deduce the classification result. We realize extensive simulations to show that ParallelNet achieves an outstanding performance in terms of recognition accuracy and robustness in severely noisy conditions.