无线电工程2024，Vol.54Issue(6)：1440-1445,6.DOI:10.3969/j.issn.1003-3106.2024.06.012
基于CNN-BiLSTM混合神经网络的雷达信号调制方式识别
Radar Signal Modulation Recognition Based on CNN-BiLSTM Hybrid Neural Network
房崇鑫 1盛震宇 1夏明 2周慧成1
作者信息
- 1. 中国船舶集团有限公司第七二四研究所,江苏南京 211153
- 2. 中国科学院大学跨学科工程研究中心,北京 100049
- 折叠
摘要
Abstract
For radar signals with time-frequency characteristics,traditional radar signal recognition methods are unable to meet the needs of accurate recognition of signal types.Therefore,it is necessary to evaluate the specific information of target radar effectively by collecting and analyzing the time-frequency characteristics in the radar signal pulse.A Convolution-Bidirectional Long Short-Term Memory(CNN-BiLSTM)hybrid neural network model is designed.The network mainly uses the time series memory characteristics of BiLSTM to deeply mine the time domain characteristics of radar signals.The weight sharing feature and the time-frequency characteristics of radar signals captured by CNN layer are combined.Then the signal characteristics of the above are used to jointly complete the identification of the radar signal modulation mode.The experimental results show that the method has high accuracy in the recognition of several radar signals,with the average value reaching 95.349％.It is better than the network using only a single feature and the traditional algorithm,and has good anti-noise ability.关键词
深度学习/卷积-双向长短时记忆混合神经网络/雷达信号调制识别Key words
deep learning/CNN-BiLSTM hybrid neural network/radar signal modulation identification分类
电子信息工程引用本文复制引用
房崇鑫,盛震宇,夏明,周慧成..基于CNN-BiLSTM混合神经网络的雷达信号调制方式识别[J].无线电工程,2024,54(6):1440-1445,6.
