基于神经网络的时频域雷达信号调制参数估计  

Modulation parameters estimation method for radar signals in time-frequency domain based on neural network

作　　者：邹兴文[1] 

机构地区：[1]西南交通大学信息科学与技术学院,四川成都610031

出　　处：《航天电子对抗》2008年第5期32-35,共4页Aerospace Electronic Warfare

基　　金：国家自然科学基金资助项目(60702026;60572143)

摘　　要：针对常规时频分析方法对多分量雷达信号分析的不足,提出一种基于神经网络的时频域多分量雷达信号调制参数估计新方法。把多分量雷达信号的时频分布作为灰度图像,通过神经网络训练得到高分辨率时频平面图,然后在时频域实现雷达信号各分量的调制参数估计,相对于时频重排等方法,其对时频分布的处理不需要关于各分量信号的先验知识。仿真实验表明,该方法在得到比时频重排方法更高分辨率时频图的基础上,能够较准确地估计各分量雷达信号调制参数。Conventional time-frequency analysis methods are exposed to some limitations in analyzing multicomponent radar signals. A method for detecting modulation parameters of multi-component radar signals is proposed based on neural network. The time-frequency distribution of multi-component radar signals is regarded as a grayscale image. Neural network training can lead to the improvement of time-frequency resolution. Radar signals modulation parameters are extracted in the time-frequency plane. Compared with time-frequency reassignment methods, it requires no priory knowledge of the signal components. Simulation results demonstrate that the method can extract effectively modulation parameters of multi-component radar signals in the highly concentrated time-frequency distribution plane.

关 键 词：雷达信号 神经网络 时频分布 调制参数估计 

分 类 号：TN957.51[电子电信—信号与信息处理] TP183[电子电信—信息与通信工程]