A New Radar Signal Recognition Method Based on Optimal Classification Atom and IDCQGA

Wan, Jian; Ruan, Guoqing; Guo, Qiang; Gong, Xue

doi:10.3390/sym10110659

Cited by 1 publication

(1 citation statement)

References 32 publications

(44 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are mainly two categories of approaches for radar intra-pulse signal modulation classification: the traditional feature extraction-based approaches and the recent deep learning-based ones. For the first category, the algorithms usually extract some useful features from the signal before classification [4][5][6][7][8][9]. The feature extraction methods employed in these algorithms include time-frequency transform using short-time Fourier transform (STFT) in [4], Choi-Williams' time-frequency distribution in [9], power feature extraction using Rihaczek distribution (RD) and Hough transform (HT) in [5], integrated quadratic phase function (IQPF) and fractional Fourier transform (FrFT) in [6], time-frequency transform using Wigner Ville distribution (WVD) and FrFT in [7], and optimal classification atom and improved double-chains quantum genetic algorithm (IDCQGA) in [8].…”

Section: Introductionmentioning

confidence: 99%

Radar Intra–Pulse Signal Modulation Classification with Contrastive Learning

et al. 2022

View full text Add to dashboard Cite

The existing research on deep learning for radar signal intra–pulse modulation classification is mainly based on supervised leaning techniques, which performance mainly relies on a large number of labeled samples. To overcome this limitation, a self–supervised leaning framework, contrastive learning (CL), combined with the convolutional neural network (CNN) and focal loss function is proposed, called CL––CNN. A two–stage training strategy is adopted by CL–CNN. In the first stage, the model is pretrained using abundant unlabeled time–frequency images, and data augmentation is used to introduce positive–pair and negative–pair samples for self–supervised learning. In the second stage, the pretrained model is fine–tuned for classification, which only uses a small number of labeled time–frequency images. The simulation results demonstrate that CL–CNN outperforms the other deep models and traditional methods in scenarios with Gaussian noise and impulsive noise–affected signals, respectively. In addition, the proposed CL–CNN also shows good generalization ability, i.e., the model pretrained with Gaussian noise–affected samples also performs well on impulsive noise–affected samples.

show abstract