摘要
目前,直方图法、PRI变换法以及平面变换法等一维、二维分选方法无法有效利用雷达信号脉冲描述字进行分选,而现有多参数聚类的多维分选方法,又存在最佳分类个数和误差范围的选择问题。针对这些问题,提出了基于小样本集推理的雷达信号多维分选技术。计算机仿真结果表明,该方法在提高分选正确率的同时可以有效降低计算复杂度。利用卫星接收到的雷达信号数据进一步验证了采用该技术的分类器具有良好的推广性。
Currently, one-dimensional or two-dimensional sorting techniques, such as histogram sorting, pulse repetition interval （PRI） conversion, and plane conversion ere, cannot make full use of the pulse- description word to classify the radar signals. As well as multidimensional sorting techniques by multi- parameter clustering cannot eomprornise between optimal category number and error span. Against them a multidimensional sorting technique of radar signals based on inference from small samples is offered. The results of computer emulation show that the technique can lower computation complexity and improve correct recognition rate. Furthermore, the radar signal data received by satellite testify the classifier presented by the technique has good generalization.
出处
《航天电子对抗》
2014年第2期54-58,共5页
Aerospace Electronic Warfare
关键词
小样本集推理
雷达信号
分选技术
inference from small samples
radar signal
multidimensional sorting
