基于改进VGG-13卷积神经网络的单脉冲信号分选

Signal sorting method of single pulse characteristic parameters based on modified VGG-13 convolutional neural network

在现有的针对脉冲描述字(Pulse Description Word,PDW)的信号预分选方法出现错误的情况下,针对脉冲重复间隔(Pulse Repetition Interval,PRI)的信号主分选的运算迭代次数和误差将会明显增加.为解决此类问题导致的整个分选结果所出现的错误.提出了基于单脉冲特性参数的信号分选方法,将信号到达时间差(Difference Time of Arrival,DTOA)特性作为新的单脉冲参数和PDW进行结合,共同完成聚类,获得更为准确的分选结果.仿真实验表明,在PDW重复度较高的情况下,会导致信号预分选出现错误,进而使最终的分选结果出现错误.而本方法在传统分选出现错误时,依然可以得到准确的雷达信号分选结果,提高了复杂环境下的分选准确率.

In the case of an error in the existing signal presorting method for pulse description word(PDW),the number of operation iterations and error of pulse repetition interval(PRI)in the main signal sorting will be significantly increased.This paper presented a signal sorting method based on single pulse parameters,time difference of concatenated(DTOA)features as a new single pulse parameters and PDW to clustering,to obtain the separation results.Simulation results showed that in the case of higher PDW repeatability,it would lead to errors in signal pre-sorting,which would lead to errors in final sorting results.However,the proposed signal sorting method based on the characteristic parameters of monopulse could still obtain accurate radar signal sorting results when the traditional sorting error occurs,which improved the sorting accuracy in complex environment.