摘要
针对单准则设计的波形难以满足雷达多状态和多任务的问题,提出了一种杂波条件下利用一维卷积神经网络的认知雷达波形设计(CRWD-1D-CNN)方法。首先,设定环境变量,并根据互信息准则和信干噪比准则来构建训练集和测试集;其次,根据数据集的一维数据形式和采样点数,设计一个包含3个卷积层、2个全连接层的1D-CNN模型;最后,使用训练集对1D-CNN模型进行训练,利用1D-CNN对一维数据之间非线性关系的学习能力来学习互信息准则和信干噪比准则,然后,使用训练后的1D-CNN生成波形。为衡量雷达综合性能,提出了一种目标最终识别率指标。实验结果表明,采用CRWD-1D-CNN方法设计的波形作为雷达发射信号时,与使用互信息准则生成的波形相比,雷达综合性能平均提升0.64%,与使用信干噪比准则生成的波形相比平均提升2.13%,证明了CRWD-1D-CNN方法可联合互信息准则和信干噪比准则,提高雷达综合性能。
A design method of cognitive radar waveform(CRWD-1D-CNN)using one-dimensional convolutional neural network(1D-CNN)in the presence of clutter is proposed to solve the problem that the waveform designed using single-criterion is difficult to satisfy the radar multi-state and multi-task.Firstly,environment variables are set,and a training set and a test set are constructed according to the mutual information(MI)criterion and the signal-to-interference and noise ratio criterion(SINR);Secondly,a 1D-CNN model containing 3 convolutional layers and 2 full connection layers is designed based on the data form and the number of sampling points;Finally,the 1D-CNN model is trained with the training set,and the learning ability of the 1D-CNN is made use of to learn the MI criterion and SINR criteria,then the trained 1D-CNN is used to generate waveform.An index of target final recognition rate and an index of target recognition rate index are proposed to measure the comprehensive performance of radar.Simulation results show that when the waveform designed by CRWD-1D-CNN method is used as radar transmitting signal,the overall performance of the radar is improved by 0.64%and 2.13%on average compared with using the waveforms generated by using MI and SINR criteria.These results prove that the CRWD-1D-CNN method can effectively combine the MI criterion and SINR criterion,and improve the comprehensive performance of radar.
作者
赵俊龙
李伟
甘奕夫
邹鲲
ZHAO Junlong;LI Wei;GAN Yifu;ZOU Kun(Information and Navigation College,Air Force Engineering University,Xi’an 710077,China)
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2021年第4期69-76,共8页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金资助项目(61571456)
航空科学基金资助项目(20160196001)
陕西省自然科学基金资助项目(2020JM-347)。
关键词
波形设计
互信息准则
信干噪比准则
一维卷积神经网络
waveform design
mutual information criterion
signal to interference and noise ratio criterion
one-dimensional convolutional neural network
