| Title |
Novel Binary Time-frequency Image-based Modulation Classification for Single and Composite Modulated Radar Signal |
| Authors |
송건호(Geonho Song) ; 전강혁(Ganghyuk Jeon) ; 서동호(Dongho Seo) ; 이원진(Wonjin Lee) ; 윤동원(Dongweon Yoon) |
| DOI |
https://doi.org/10.5573/ieie.2024.61.6.3 |
| Keywords |
Radar signal; Modulation classification; Composite modulated radar signal; Time-frequency image; Deep learning |
| Abstract |
In modern warfare, as numerous weapon systems adopt electromagnetic signals to cope with expanding battlefield areas, the importance of electromagnetic warfare is becoming increasingly prevalent. In electromagnetic warfare, analyzing the radar signals of hostile forces is a crucial activity to gain superiority. Particularly, the technique to classify modulation schemes of radar signals collected without prior information plays a vital role in enabling friendly forces to identify the presence of threats and obtain valuable intelligence. For radar signal modulation classification, the use of binary time-frequency image (TFI) has the advantage of improving system memory efficiency over grayscale TFI, but with the drawback of image pattern distortion during binarization. This paper proposes a novel binary TFI generation method that reduces pattern distortion. To achieve this, we first generate a grayscale TFI of the received radar signal, then reduce the noise in the image by extracting the frequency component-related information through temporal marginalization of the grayscale TFI. Subsequently, the noise-reduced image is standardized and clipped along the frequency axis for pattern enhancement, followed by global thresholding for binarization. Through computer simulations, we show that the binary TFI generated with the proposed method can effectively classify 37 single and composite radar modulation schemes. Furthermore, we verify that for various deep learning models, the proposed method shows better classification performance than the binary TFIs generated using conventional methods. |