The increasingly complex radio environment may cause the received low probability of intercept (LPI) radar signals to overlap in time–frequency domains. Analyzing overlapping LPI radar signals requires identifying the modulation type and estimating the parameters of each component. Prior research performs overlapping signal analysis as a multistage task, where each stage is designed to perform a part of the task. The multistage system will increase the calculation burden and cannot be optimized as a whole. Instead, this article proposes a novel framework for analyzing overlapping signals in a single stage. Specifically, we develop a joint semantic learning deep convolutional neural network (JSLCNN) that jointly learns three tasks, i.e., feature restoration, modulation classification, and parameter regression. Since the whole cognitive pipeline is a single network, it can be optimized end-to-end directly on cognitive performance. To verify the validity of the proposed JSLCNN, numerous comparative experiments are carried out in terms of modulation recognition and parameter estimation of overlapping signals. Experimental results demonstrate that the JSLCNN has desirable extensibility for identifying unseen signal combinations and robustness against unknown jamming. Furthermore, we show that the JSLCNN outperforms other existing approaches in generic real-time parameter estimation for LPI radar signals.

Automatically recognizing the modulation of radar signals is a necessary survival technique in electronic intelligence systems. In order to avoid the complex process of the feature extracting and realize the intelligent modulation recognition of various radar signals under low signal-to-noise ratios (SNRs), this paper proposes a method based on intrapulse signatures of radar signals using adaptive singular value reconstruction (ASVR) and deep residual learning. Firstly, the time-frequency spectrums of radar signals under low SNRs are improved after ASVR denoising processing. Secondly, a series of image processing techniques, including binarizing and morphologic filtering, are applied to suppress the background noise in the time-frequency distribution images (TFDIs). Thirdly, the training process of the residual network is achieved using TFDIs, and classification under various conditions is realized using the new-trained network. Simulation results show that, for eight kinds of modulation signals, the proposed approach still achieves an overall probability of successful recognition of 94.1% when the SNR is only −8 dB. Outstanding performance proves the superiority and robustness of the proposed method.

The recognition and estimation of frequency-modulated continuous-wave (FMCW) radar signals is critical for both military electronic countermeasures (ECM) and civilian autonomous driving. However, the increasingly complex radio environment poses two new challenges for detection equipment. First, multiple FMCW radars can share the same frequency band, resulting in overlap of received FMCW signals in the time-frequency domain. Second, unexpected signals in unknown spectrum environments can affect the cognitive performance of FMCW signals. This paper proposes a semantic-based learning network (SLN) that simultaneously learns modulation classification and parameter regression of FMCW signals. By integrating recognition and estimation into a single network, the system can be optimized end-to-end as a whole. Additionally, instance-level semantic learning facilitates the parallel analysis of multiple components in overlapping signals. Finally, contrastive clustering in SLN achieves suppression of unexpected signals. Numerous comparative experimental results demonstrate that SLN has the desirable ability to simultaneously recognize and estimate FMCW signals in real-time, even in unknown spectrum environments.

The emergence of active jamming technology has led to a surge of interest in jamming suppression. This work proposes a one-dimensional adaptive multi-scale residual autoencoder (1-DAMRAE) to take overlapping signals (echo signals and jamming signals) as input and restore correspondingly-sized echo signals, which frames the jamming suppression as a regression problem. 1-DAMRAE incorporates three key components. First, adaptive multi-scale convolutional blocks are developed to adaptively fuse multi-scale information for a more robust representation learning. Second, a fully residual design is explored to improve information flow within and across blocks. Third, given the unknown of jamming sources, our method encodes contextual dependencies and produces a plausible hypothesis for the potential missing information to handle out-of-distribution (OOD) jamming. To verify the reliability of 1-DAMRAE, the dynamic short-range detection system model is established. Experimental results show that 1-DAMRAE achieves excellent anti-jamming performance while keeping complexity and computation cost at a low level. Even facing unseen OOD jamming, 1-DAMRAE can still achieve desirable anti-jamming performance. The model and data are available in https://github.com/Drchenjust/1-DAMRAE .