一种基于Fast ICA和K-Means++融合算法的雷达脉冲聚类方法  

A Method of Radar Pulse Clustering Based on Fusion Algorithm of Fast ICA and K-Means++

作　　者：李伟[1] LI Wei(The 723 Institute of CSIC,Yangzhou 225101,China)

机构地区：[1]中国船舶重工集团公司第七二三研究所,江苏扬州225101

出　　处：《舰船电子对抗》2021年第4期61-64,69,共5页Shipboard Electronic Countermeasure

摘　　要：为在辐射源非协作的情况下对脉冲流密度很大的雷达信号进行有效分选,要克服Fast ICA算法在分选雷达脉冲时分离速度慢且实现复杂、K-Means++算法在分选雷达脉冲描述字PDW时对噪声和孤立噪点敏感且需要先验条件K值的缺点。提出了一种在CPU+GPU异构系统上实现的这2种算法的融合方案,融合后的算法能有效克服单独利用一种算法的缺点、综合两者的优点。仿真实验表明提出的融合方案能有效提升雷达信号分选的准确性。For effectively sorting the radar signals with high pulse flow density in the case of non-cooperative emitters,it is necessary to overcome the shortcomings:slow separation speed and complex implementation when sorting radar signal pulse by using fast ICA algorithm,being sensitive to noise and isolated noise point and requiring a priori condition K value when sorting pulse description word(PDW)information by using K-Means++algorithm.A fusion scheme of the two algorithms implemented on the CPU+GPU heterogeneous system is proposed.The fused algorithm can effectively overcome their respective shortcomings and combine the advantages of the two.Simulation experiments also show that the proposed fusion scheme can effectively improve the accuracy of radar signal sorting.

关 键 词：独立成分分析 聚类 雷达信号分选 算法融合 

分 类 号：TN971.1[电子电信—信号与信息处理]