基于深度学习的复杂环境下无线电信号自动检测方法研究

打开文本图片集

中图分类号：TN927.5；TN98 文献标志码：B 文章编码：1672-7274（2026）02-0023-04

Research on Automatic Detection of Radio Signals in Complex Environments Based on Deep Learning

LIYuxuan，FANWei （TheStateRadio_monitoring_center TestingCenter （SRTC），Beijing10o041，China）

Abstract： To addressthe challenge of detecting burst radio signals in complex electromagnetic environments， this paper proposes a deep learning-based detection method using a time-frequency dual-path convolutional network. By incorporatingIQdata augmentation，atention mechanisms，and structured pruning，the model enhances detection performance while achieving lightweight deployment.Experimental results demonstrate that the proposed method maintains ahigh detection probability evenunderlowsignal-to-noiseratio conditions （-12 dB），significantly improving robustness and real-time capability compared to traditional approaches.

Keywords： complex electromagnetic environment; radio signal detection;deep learning; time-frequency dualpath network; model lightweighting

0 引言

传统能量检测和特征工程方法在低信噪比条件下常常面临检测灵敏度的瓶颈，难以满足现代通信保障感知需求[1]。（剩余4554字）