太赫兹科学与电子信息学报2026，Vol.24Issue(7)：863-868,6.DOI:10.11805/TKYDA2025159
信号快速识别及应用
Rapid signal detection and application
孟祥峰 1刘军 1王元恺 1杨青1
作者信息
- 1. 中国电子科技集团公司 第四十一研究所,山东 青岛 266555
- 折叠
摘要
Abstract
With the ever-increasing proliferation of frequency-using devices,the electromagnetic environment has become exceptionally complex.In addition to authorized frequency users,unauthorized ones also exist in the electromagnetic space,making the detection and identification of these signals of great significance.Manual screening for anomalous interference signals is extremely difficult,and traditional identification methods have limited capabilities that preclude accurate recognition.To address this,we propose a deep learning-based signal identification method to facilitate the discovery of anomalous interference signals.Signal spectral segments are extracted from the wideband spectrum,and signal features are computed by a deep learning model.These features are then matched against those in a signal feature database via similarity matching,ultimately returning the identification label and confidence score.For unknown signals not previously added to the feature library,the model returns"unknown"or outputs an extremely low identification confidence.Meanwhile,the proposed scheme supports online,dynamic adding of signal labels.Benefiting from the model's flexible label management mechanism,it offers high usability and maintains good performance even in open environments.Furthermore,the model demonstrates high identification efficiency;practical verification shows that the algorithm achieves an identification speed exceeding 1 000 signals per second.关键词
深度学习网络/宽带频谱/相似度匹配/标签管理/动态添加Key words
deep learning/broadband spectrum/similarity matching/label management/dynamic adding分类
信息技术与安全科学引用本文复制引用
孟祥峰,刘军,王元恺,杨青..信号快速识别及应用[J].太赫兹科学与电子信息学报,2026,24(7):863-868,6.
