基于AdaBoost分类器的典型战术数据链信号识别方法

A recognition method of classical tactical data links based on AdaBoost classifier

现代作战体系主要依靠战术数据链传递战术信息,因此对侦收到的敌方数据链信号进行准确高效识别的重要性日益凸显。现有特征提取方法是基于单一的信号维度,无法在更高维度条件下获得信号多维耦合特征,同时单一分类器分类识别概率不高。针对此问题,提出在数据链信号时频图和循环谱图中提取多维特征,采用基于AdaBoost算法的自适应优化组合分类器,形成AdaBoost强分类器对数据链信号进行识别。通过仿真分析可知,基于多维特征的AdaBoost强分类器对数据链信号的分类识别,相较于单个分类器,其识别准确率有明显提升,具有良好的应用前景。

In modern warfare,combat units mainly rely on tactical data links to exchange tactical messages,so the importance of accurately and efficiently identifying enemy tactical data link signals is becoming prominent.Current feature extraction methods are not capable of extracting coupling features of signal in the condition of multi-dimension,because they extract feature in single dimension only.Meanwhile,single classifier could not reach high recognition accuracy.In terms of questions above,an approach that constructs combined multiple clas⁃sifiers and adjusts weight dynamically based on AdaBoost is proposed,after extracting multi-dimensional features of signals from their time-frequency images and cyclic spectrum images.Simulation analysis shows that proposed approach could improve recognition accuracy clearly compared with single classifier,and has good application prospect.