Deep learning-based LPI radar signals analysis and identification using a Nyquist Folding Receiver architecture

Wan, Tao; Jiang, Kaili; Ji, Hao; Tang, Bin

doi:10.1016/j.dt.2021.09.019

Cited by 6 publications

(4 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some methods use the wavelet packet transform (WPT) to extract the features and then a probabilistic support vector machines (SVM) to implement the radar emitter recognition [35][36][37]. To improve the performance of the signal extraction, deep learning uses multiple hidden layers perceptron to guarantee a more effective in the extraction of the signal, see for istance [38][39][40][41][42][43][44].…”

Section: A Fifth Block-off-line Training-work Off-line To Train the C...mentioning

confidence: 99%

On the anti‐intercept features of noise radars

Galati,

Pavan

2024

IET Radar Sonar & Navi

View full text Add to dashboard Cite

Robustness against Electronic Warfare/Electronic Defence attacks represents an important advantage of Noise Radar Technology (NRT). An evaluation of the related Low Probability of Detection (LPD) and of Intercept (LPI) is presented for Continuous Emission Noise Radar (CE‐NR) waveforms with different operational parameters, that is, “tailored”, and with various “degrees of randomness”. In this frame, three different noise radar waveforms, a phase Noise (APCN) and two “tailored” noise waveforms (FMeth and COSPAR), are compared by time–frequency analysis. Using a correlator (i.e. a two antennas) receiver, assuming a complete knowledge of the band (B) and duration (T) of the coherent emission of these waveforms, it will be shown that the LPD features of a CE‐NR do not significantly differ from those of any CE radar transmitting deterministic waveforms. However, in real operations, B and T are unknown; hence, assuming an instantaneous bandwidth estimation will show that the duration T can be estimated only for some specific “tailored” waveforms (of course, not to be operationally used). The effect of “tailoring” is analysed with prospects for future work. Finally, some limitations in the classification of these radar signals are analysed.

show abstract

Section: A Fifth Block-off-line Training-work Off-line To Train the C...mentioning

confidence: 99%

On the anti‐intercept features of noise radars

Galati,

Pavan

2024

IET Radar Sonar & Navi

View full text Add to dashboard Cite

show abstract

“…Correspondingly, These methods have been shown to achieve high accuracy (up to 90%) in classifying radar emitter signals even in environments with high electromagnetic density and noise. For examples, the classic open-source CNN models including GoogleNet [20], ResNet [21] and MobileNetV2 [22] have achieved remarkable performance in the classification of timefrequency images due to its strong generalization abilities.…”

Section: Introductionmentioning

confidence: 99%

Dilated CNN Design Approach for Extracting Multi-Scale Features in Radar Emitter Classification

Guo,

Wu,

Guo

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Radar emitter classification plays an increasingly significant role in the electronic reconnaissance system. Due to many convolutional neural network (CNN)-based approaches suffer from insufficient spatial receptive fields and inadequate feature representation, the classification accuracy is poor in low signal-to-noise ratio (SNR) conditions. Therefore, in this paper, we stress the importance of multiscale dilated convolutions for target feature extraction, and propose two novel CNN architecture design approaches called multi-scale dilated residual network (MDRN). By combining multi-scale dilated convolutions with residual architecture, MDRN not only has a larger receptive field, but also can learn more diverse features, thereby improving the ability to process time-frequency images (TFI) under highnoise energy conditions. Moreover, compared with the original residual model, MDRN does not increase any parameter complexity or floating-point operations per second (FLOPS). Experiments on the TFI classification task show that the proposed MDRN has superior performance over state-of-the-art CNN models.INDEX TERMS intra-pulse modulation classification; dilated convolution; feature maps; multi-scale feature fusion; time-frequency image

show abstract

“…Although the NYFR can provide a hundred percent probability of intercept (POI) [18], the existing signal processing methods are mainly based on the CS or template matching method. The sparse signal reconstruction via CS-based methods or other optimization methods has high computational and complexity [21]. And there are high signal-tonoise (SNR) requirements for the information acquisition [22].…”

Section: Introductionmentioning

confidence: 99%