基于类别距离和Bhattacharyya距离的雷达信号特征评价

摘要

This paper presented a novel evaluation criterion of optimal feature by means of comparative analysis of the instantaneous autocorrelation method, wavelet transform and time-frequency atom method. It chosen the features extracted in different algorithms by the distance between classes, within-class distance and Bhattacharyya distance with respect to the feature distribution and upper bound of the error rate. The simulation results show that the evaluation criterion of optimal feature is effective. It provides a valuable reference for the feature evaluation of the radar emitter signal.%从瞬时自相关法、小波变换法和时频原子法提取的脉内特征比较分析入手,提出了一种新的最优特征评价准则.以类内距离、类间距离和Bhattacharyya距离为基础,从特征的空间分布和错误识别率的上界等方面对不同算法提取出的脉内特征进行分析,实现了最优特征的选择.实验的仿真结果表明,这种最优特征评价准则是有效的,为雷达辐射源信号的特征评价提供了有意义的参考.