摘要
针对现有实时分选方法依赖滑动窗(单次处理脉冲数)的选取、分选效率低以及分选结果存在严重增批现象等缺陷,将脉冲串的接收视为数据流过程,从而设计了一种基于进化数据流聚类的雷达信号实时分选框架。该框架分为在线处理和离线分析两个阶段,通过引入时态密度特征避免了主观上对滑动窗长度的选取,并利用衰变检测和噪声点检测来提高在线聚类的效率。离线阶段通过对历史快照的分析可以判明雷达的活动情况,并将属于一部雷达的脉冲批组进行关联。仿真实验表明了该框架的有效性和可行性。
In view of the existing real time sorting methods having defects of relying on sliding window （the pulse number to be conducted one time）, low sorting efficiency and serious phenomenon of increased number, pulse sequence is regarded as a data flow process, thus a reabtime signal sorting framework based on online clustering of evolution data stream is designed. The framework is divided into two stages as an online part and an off[ine analysis part. With the utilization of temporal density characteristics, the subjective selection of sliding window length is avoided. And the disintegration detection and noise detection are used to improve the efficiency of online clustering. Off-line stage with the analysis of the historical snapshot can determine the movements of the radar, and merge groups of pulses which come from the same emitter. The simulation experiment shows the validity and feasibility o^f the framework.
出处
《航天电子对抗》
2016年第2期6-9,17,共5页
Aerospace Electronic Warfare
关键词
信号分选
数据流
时态密度
放射传播聚类
离线分析
signal deinterleaving
data stream
temporal density
affinity propagation cluster
off line analysis
