Augmentation Algorithms of Radar Signals Dataset via Wigner-Ville Images.

Saved in:

Bibliographic Details
Title:	Augmentation Algorithms of Radar Signals Dataset via Wigner-Ville Images.
Authors:	Zhao, Dazhi¹ zhaodazhimail@163.com, Huang, Lei¹ 1437216871@qq.com, Zhou, Hui¹ 1048345246@qq.com, Zhang, Zelin² 20170020@huat.edu.cn
Source:	Engineering Letters. May2026, Vol. 34 Issue 5, p1469-1477. 9p.
Subjects:	Data augmentation, Radar signal processing, Signal-to-noise ratio, Time-frequency analysis, Deep learning
Abstract:	Deep learning for radar signal classification is constrained by scarce labeled data and high acquisition costs. To address this, radar waveforms--including Linear Frequency Modulation (LFM), Rectangular (Rect), and Barker--are transformed into time-frequency images via the Wigner-Ville Distribution (WVD) and augmented using four image-based methods: Random Erasing (RE), Histogram Equalization (HE), Grayscale Adjustment (GA), and Image Inversion (II). In clean conditions, augmentation improves accuracy, with higher structural similarity (SSIM) correlating with stronger gains, while severe distortion degrades performance. Under varying signal-to-noise ratios, all methods show performance decline at lower SNRs--more pronounced for low-SSIM methods--with recovery toward clean-condition levels at higher SNRs. These results confirm that preserving time-frequency structural fidelity is critical in both clean and noisy environments. The approach effectively mitigates data scarcity and demonstrates cross-domain applicability of image-based augmentation in radar signal processing. [ABSTRACT FROM AUTHOR]
	Copyright of Engineering Letters is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	Deep learning for radar signal classification is constrained by scarce labeled data and high acquisition costs. To address this, radar waveforms--including Linear Frequency Modulation (LFM), Rectangular (Rect), and Barker--are transformed into time-frequency images via the Wigner-Ville Distribution (WVD) and augmented using four image-based methods: Random Erasing (RE), Histogram Equalization (HE), Grayscale Adjustment (GA), and Image Inversion (II). In clean conditions, augmentation improves accuracy, with higher structural similarity (SSIM) correlating with stronger gains, while severe distortion degrades performance. Under varying signal-to-noise ratios, all methods show performance decline at lower SNRs--more pronounced for low-SSIM methods--with recovery toward clean-condition levels at higher SNRs. These results confirm that preserving time-frequency structural fidelity is critical in both clean and noisy environments. The approach effectively mitigates data scarcity and demonstrates cross-domain applicability of image-based augmentation in radar signal processing. [ABSTRACT FROM AUTHOR]
ISSN:	1816093X