基于SOFM网络聚类雷达信号分选预处理改进算法  

An improved pre-processing algorithm of radar signal sorting based on SOFM clustering

作　　者：郑子扬[1] 陈永游[1] 张君[1] 史敏[1] 贺刚[1] 

机构地区：[1]中国航天科工集团8511研究所,江苏南京210007

出　　处：《航天电子对抗》2013年第3期42-45,50,共5页Aerospace Electronic Warfare

摘　　要：采用自组织特征映射神经网络将含有噪声背景的多雷达混合脉冲描述字(PDW)数据进行聚类,根据距离特征进行噪声脉冲滤除。在密度可分的前提下,依据密度特征进行再聚类处理,最后输出各类的脉冲集合。仿真实验结果表明,该方法可以有效分离复杂电磁环境下的雷达脉冲信号,对捷变频雷达信号具有良好的预处理效果。With Self-Organizing Feature Mapping （SOFM） neural networks, the data of Pulse Description Word （PDW） of multi-radar is clustered in noisy background, then noise pulses are filtered based on distance feature. On the assumption that density is separable, the second clustering process is implemented according to density feature. In the end,pulse sets of different clusters are output. Results of simulation experiment indicate that radar pulses are effectively separated in complex electromagnetic environment, and the method is also suitable to pre-processing for frequency-agile radar.

关 键 词：自组织特征映射 神经网络 信号分选 

分 类 号：TN957.51[电子电信—信号与信息处理]