基于神经网络的高频地波雷达目标到达角估计  

DOA Estimation Based on Neural Network for HFGWR

作　　者：严颂华[1] 吴世才[1] 吴雄斌[1] 

机构地区：[1]武汉大学电子信息学院,武汉430079

出　　处：《电子与信息学报》2008年第2期339-342,共4页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(60571065);国家863计划(20001AA631050)

摘　　要：该文利用神经网络进行高频地波雷达目标到达角估计。论文分别采用RBFN和GRNN构造了基于函数逼近和模式编码的到达角估计网络,介绍了网络结构、数据仿真的过程和应用于高频地波雷达目标定向的实际效果。数据仿真和现场实验的分析结果表明基于模式编码的GRNN网络到达角估计方法鲁棒性较好,在低信噪比时能够给出正确估计。Direction of Arrival （DOA） technology based on neural network is discussed. In the paper, the DOA method based on function approach and model classification with coding is presented and it employs two kinds of neural networks： Radial Basis Function Network（RBFN） and General Regression Neural Net （GRNN）. The paper introduces the network structure, simulation and the application to High Frequency Ground Wave Radar（HFGWR）. The simulation and real data processing verifies that the model classification method based on GRNN offers better performance than others, its performance is good even the signal-to-noise of the signal is low until 4 dB.

关 键 词：神经网络 到达角估计 模式分类 高频地波雷达 

分 类 号：TN958[电子电信—信号与信息处理]