基于SpecUNeSt 的无线电信号一维检测

打开文本图片集

关键词： 信号检测； 分割网络； SpecUNeSt； 特征提取； 信号频谱； 分裂注意力机制中图分类号：TN929.5⁃34 文献标识码： A 文章编号：1004⁃373X（2025）12⁃0041⁃06

Abstract：Signal detection is an important part of electromagnetic space reconnaissance and perception， and the traditional detection algorithm is difficult to detect relatively weak electromagnetic signals in complex electromagnetic countermeasure environments. On the basis that the intelligent technology has unique advantages in target detection and recognition， a radio signal one⁃dimension detection algorithm based on SpecUNeSt is proposed by researching on the intelligent signal detection and according to the characteristics of the signal spectrum. This algorithm has the advantage of overcoming the degradation problem of deep learning networks， which can accurately estimate signal parameters by means of the excellent feature extraction ability and split attention mechanism of the segmentation network： UNet， and using signal spectra under different bandwidths and signal⁃ to⁃noise ratios as features. The split attention mechanism in the model can improve the connection between signal features across channels and effectively reduce the impact of noise on signal detection. The simulation results show that the radio signal one ⁃ dimensional detection algorithm based on SpecUNeSt can enhance the anti⁃inference ability compared with traditional algorithm， and can improve the detection accuracy and recall rate compared with one⁃dimensional UNet detection.

Keywords： signal detection; segmentation network; SpecUNeSt; feature extraction; signal spectrum; split attention mechanism

0 引 言

在干扰比较少、电磁环境干净的情况下，传统信号处理手段依然可以发挥较大的作用，但是对于干扰对抗等复杂电磁对抗环境，该方法则难以高效检测出相对弱小的电磁信号。（剩余6618字）