摘要
提出了一种基于模式编码进行宽带信号到达角估计的方法。该方法通过广义回归神经网络(GRNN)对信号的线性时频协方差矩阵进行分类来实现。介绍了网络结构,数据仿真和应用于高频地波雷达目标定向的实际效果。特别介绍了通过窄带信号数据训练网络然后应用于宽带信号的训练方法,适于宽带训练数据难以获得的情形。仿真和实测数据的分析结果显示该估计方法能适应阵列的误差模式且鲁棒性好,在较低信噪比时能够给出正确估计。
A Direction of Arrival (DOA) technology based on model classification by coding are presented,it employ General Regression Neural Net (GRNN) to classify signals by their linear-time-frequency correlation matrices.This paper introduce the structure of the network,data simulation and application in High Frequency Ground Wave Radar(HFGWR).It also give the method to train the neural network by narrow-band signals,which overcome the problem of getting the wide-band training data.The simulation and real data processing verified its adaptability to antenna errors and robust at low signal-to-noise(SNR).
出处
《火力与指挥控制》
CSCD
北大核心
2010年第5期40-43,共4页
Fire Control & Command Control
基金
国家"863"基金(2006AA09A303)
大学生国家创新计划基金资助项目(2006043)