一种深度学习的雷达辐射源识别算法
发布时间：2018-10-20 19:47
【摘要】：针对传统依靠于人工经验提取雷达辐射源特征方法的不足,提出了一种新颖的基于联合深度时频特征的辐射源识别算法.首先将时域信号变换到二维时频域,并利用随机投影和主成分分析方法分别从维持子空间和能量角度对时频图像降维;接着在预训练阶段,利用无标签的样本信号层级训练深度模型,再根据类别信息精调网络参数;最后,构造了逻辑回归分类来完成识别任务.仿真实验中利用6种辐射源信号验证了提出算法的有效性,结果表明,联合深度特征更加有助于提高识别准确度,算法运行更加高效.
[Abstract]:Aiming at the shortcoming of the traditional method of extracting radar emitter features based on artificial experience, a novel emitter recognition algorithm based on joint depth time-frequency features is proposed. Firstly, the time domain signal is transformed into two dimensional time-frequency domain, and then the dimension of the time-frequency image is reduced from the maintenance subspace and the energy angle by the methods of stochastic projection and principal component analysis, and then in the pre-training stage, Using the untagged sample signal level training depth model, the network parameters are carefully adjusted according to the category information. Finally, the logical regression classification is constructed to complete the recognition task. The effectiveness of the proposed algorithm is verified by using six emitter signals in simulation experiments. The results show that the combined depth feature is more helpful to improve the recognition accuracy and the algorithm runs more efficiently.
【作者单位】： 海军工程大学电子工程学院;
【基金】：国家自然科学基金资助项目(61501484) 国家“863”高技术研究发展计划资助项目(2014AA7014061)
【分类号】：TN957.51
[Abstract]:Aiming at the shortcoming of the traditional method of extracting radar emitter features based on artificial experience, a novel emitter recognition algorithm based on joint depth time-frequency features is proposed. Firstly, the time domain signal is transformed into two dimensional time-frequency domain, and then the dimension of the time-frequency image is reduced from the maintenance subspace and the energy angle by the methods of stochastic projection and principal component analysis, and then in the pre-training stage, Using the untagged sample signal level training depth model, the network parameters are carefully adjusted according to the category information. Finally, the logical regression classification is constructed to complete the recognition task. The effectiveness of the proposed algorithm is verified by using six emitter signals in simulation experiments. The results show that the combined depth feature is more helpful to improve the recognition accuracy and the algorithm runs more efficiently.
【作者单位】： 海军工程大学电子工程学院;
【基金】：国家自然科学基金资助项目(61501484) 国家“863”高技术研究发展计划资助项目(2014AA7014061)
【分类号】：TN957.51
【相似文献】
相关期刊论文 前10条
1 关欣,何友,衣晓;一种新的雷达辐射源识别算法[J];宇航学报;2005年05期
2 陈根华,杨绍全;特定辐射源识别方法的研究[J];航天电子对抗;2005年02期
3 李楠;曲长文;苏峰;平殿发;;灰色关联理论在辐射源识别中的应用[J];系统仿真学报;2009年24期
4 黄渊凌;郑辉;万坚;;特定辐射源识别的信息论描述[J];西安电子科技大学学报;2012年06期
5 纪彦星;曹俊纺;陈建军;孟晓琳;;基于精细分析的辐射源识别软件系统设计及实现[J];雷达与对抗;2010年02期
6 姚s，
本文编号：2284203
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2284203.html
