Search

Highlights • A new two channel Convolutional Neural Network combining with Bi-directional Long Short-Term Memory achieves excellent performance for modulation recognition and protocol recognition. • Transfer learning is firstly introduced in waveform recognition filed to solve the model transferability problem across different recognition tasks, which can efficiently save the 60% training time and 50% training data. • A complete protocol signal dataset is provided for public research about waveform recognition in passive radar. Graphical abstract Abstract Passive radar capable of recognizing illumination of opportunities can improve the detection performance on account of its functional properties of environment adaptivity. Waveform recognition approaches based on Deep Learning can outperform traditional methods based on hand-crafted feature as shown in recent studies. In this paper, we propose a novel transferred deep learning waveform recognition method which makes use of multi-scale convolution and temporal dependency characteristics to improve the recognition performance. Firstly, we develop a two-channel convolutional neural networks combining with Bi-directional Long Short-Term Memory (TCNN-BL) architecture to extract features of different scales and merge past and future states. Then in order to solve the transferability problem across various sampling frequencies, we present a parameter transfer approach which initializes target domain classifier using source domain parameters. Based on our experiments on both public datasets and our own datasets, it can be demonstrated that the proposed approach significantly outperforms the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

• Active antagonism requires the radar to take measures in advance to avoid being jammed, which can address the challenge of mainlobe jamming. • A subpulse-level frequency agile radar and a jammer that works in a transmit/receive time-sharing mode are considered. • Under the established model, Markov Decision Process is used to describe the relationship of them. • The anti-jamming strategies can be designed based on advanced deep reinforcement learning. Among different jamming techniques, mainlobe jamming is difficult to deal with for the radar and traditional passive anti-jamming methods are less effective because the angular separation between the jammer and the target is almost the same. In contrast to these passive methods, active antagonism requires the radar to take measures in advance to avoid being jammed and this can be achieved via frequency agile (FA) radar. In order to enable the FA radar to combat the jammer and obtain good performance, a deep reinforcement learning (RL) based anti-jamming strategy design method is proposed in which a transmit/receive time-sharing jammer may adopt multiple different jamming strategies. To combat the individual jamming strategy, we propose a specialized strategy learning algorithm that treats probability of detection as the reward signal and uses proximal policy optimization to solve the RL problem of the radar and the jammer. Based on the learned specialized strategies, policy distillation technique is applied to design a unified strategy which enables the FA radar to combat multiple jamming strategies. Simulation results show that the FA radar can avoid being jammed and obtain a high probability of detection whether the jammer adopts individual or multiple jamming strategies through the proposed method. [ABSTRACT FROM AUTHOR]