| 摘要: |
| 本文提出一种基于自注意力对抗自编码器的改进雷达信号智能分选方法，旨在解决现有的基于卷积降噪自编码器的雷达信号分选方法在辐射源PRI模式存在抖动PRI时分选效果下降的问题。通过在卷积降噪自编码器中间插入自注意力机制模块提高网络特征提取能力，并加入判别器构成对抗自编码器提高模型泛化能力，显著改善了网络对存在抖动PRI的到达时间序列的分选性能。该方法在分选时将目标序列以外的信号视为异常，将混叠到达时间序列进行二进制编码后输入训练后的网络，实现对目标序列的分选提取。实验结果表明，当脉冲序列存在抖动PRI时，本文方法在理想条件以及考虑虚假脉冲和脉冲丢失等复杂电磁环境下，分选效果均优于基于卷积降噪自编码器的分选方法。 |
| 关键词: 雷达信号分选 自注意力机制 对抗自编码器 到达时间序列编码 |
| DOI： |
| 分类号:TN971 |
| 基金项目: |
|
| An Intelligent Sorting Method of Radar Signal Based on Self-attention Adversarial Autoencoder |
|
|
|
|
| Abstract: |
| In this paper, an improved radar signal intelligent sorting method based on self-attention adversarial autoencoder is proposed, which aims to solve the problem that the separation effect of the existing radar signal sorting method based on convolutional denoising autoencoder decreases when jitter PRI mode exists. By inserting the self-attention mechanism module in the middle of the convolutional denoising autoencoder to improve the network feature extraction ability, and adding a discriminator to form an adversarial autoencoder to improve the generalization ability of the model, the network significantly improves the sorting performance of the TOA series with jitter PRI. In this method, the signals other than the target sequence are regarded as anomalies during sorting, and the aliasing arrival time series is binary encoded and input into the trained network to realize the sorting and extraction of the target sequence. Experimental results show that the proposed method is better than the sorting method based on convolutional noise reduction autoencoder under ideal conditions and in complex electromagnetic environments such as spurious pulses and pulse loss when there is jitter PRI in the pulse sequence. |
| Key words: radar signal sorting self-attention mechanism adversarial autoencoder TOA series coding |