| 摘要: |
| 针对复杂战场电磁环境下的低截获概率（Low Probability of Intercept,LPI）雷达信号识别问题，提出了一种基于小波变换和多尺度残差网络的识别方法。首先利用离散小波变换提取LPI雷达信号的时频特征；然后利用多尺度残差网络的多层特征提取网络对输入信号进行深层自主学习，以获取原信号时频数据的分布式特征，从而实现对雷达信号的识别预测。在不同信噪比条件下对不同调制样式信号进行实验，证明了所提方法的可行性与有效性，以及较高的识别率和泛化能力。 |
| 关键词: 低截获概率雷达 信号识别 多尺度残差网络 小波变换 |
| DOI： |
|
| 基金项目:国家自然科学基金资助项目（61571462） |
|
| LPI radar signal recognition based on multi-scale residual network and wavelet transform |
| LIU Ying,TIAN Runlan,DONG Huixu |
| (School of Aviation Operations and Services,Aviation University of Air Force,Changchun 130022,China) |
| Abstract: |
| For the problem of low probability of intercept(LPI) radar signal recognition in complex battlefield electromagnetic environment,a recognition method based on wavelet transform and multi-scale residual network is proposed.Firstly,the time-frequency characteristics of LPI radar signals are extracted by discrete wavelet transform,and then the multi-level feature extraction of multi-scale residual network is used to deeply autonomously learn the input signals to obtain the distributed characteristics of the original signal data,thus realizing the recognition and prediction of radar signal.Experiments on different modulation modes in different signal-to-noise ratio conditions verify the feasibility and effectiveness of the proposed method,as well as a high recognition rate and generalization ability. |
| Key words: low probability of intercept radar signal identification multi-scale residual network wavelet transform |
《电讯技术》编辑部 