基于长短时记忆网络的雷达波形设计

针对单准则设计的波形难以满足雷达多工作模式和多任务问题,联合互信息(mutual information, MI)准则和信杂噪比(signal to clutter and noise ratio, SCNR)准则,提出一种基于长短时记忆(long short-term memory, LSTM)网络的雷达波形设计方法。首先，设计了由输入层、LSTM层和输出层构成的LSTM网络。其次，将基于MI准则和SCNR准则生成的信号与相应环境信息组成训练集,训练LSTM网络。最后,使用训练完成的LSTM网络设计波形,并将所提方法生成波形的雷达性能与单准则设计的波形进行对比。为衡量雷达综合性能,提出一种新的雷达综合性能指标和目标识别率。仿真结果表明,所提方法生成信号作为发射信号时,与MI准则生成信号相比,雷达综合性能平均提升0.67%,与SCNR准则相比,平均提升1.47%,证明了该方法可兼具MI准则和SCNR准则优点,提升雷达综合性能。

Waveform design of radar based on long short-term memory network

Considering the problem that the waveform designed using single-criterion is difficult to meet the radar’s multi-working mode and multi-tasking,a waveform design method based on long short-term memory(LSTM)network is proposed,through combining mutual information(MI)criterion and signal to clutter and noise ratio(SCNR)criterion.Firstly,a LSTM network composed of input layer,LSTM layer,and output layer is designed.Secondly,a train set composed of signals generated through MI criteria and SCNR criteria and the corresponding environmental information is constructed,which is used to train the LSTM network.Finally,the trained LSTM network is used to design waveform,and the radar performance of the waveform generated by the proposed method is compared with the single-criterion design waveform.In order to measure the comprehensive performance of radar,a novel comprehensive performance indicator and target reco gnition rate of radar are proposed.Simulation results show that when the signal generated by the proposed method is used as the transmit signal,the comprehensive performance of radar is increased by 0.67%on average compared with MI criterion and 1.47%compared with SCNR criterion.It is proved that the proposed method can combine the advantages of the MI criterion and SCNR criterion,and improve th e comprehensive performance of radar.