基于特征矢量输入的神经网络测向方法

提出了一种新的基于特征矢量的采集输入数据方法，经该方法训练的径向基函数神经网络（RBFNN）可用于码多分址（CDMA）系统中多源信号波达角（DOA）的估计。该方法对信道噪声不敏感，能以较少的训练样本就可得到推广能力较好的神经网络。仿真结果表明，以新方法训练的RBFNN对多源信号DOA估计精度较高，实时性好，适用于CDMA通信系统的高分辨率测向。

Direction of Arrival Estimation Method with Eigenvector- Based Radial Basis Function Neural Network

A new direction of arrival (DOA) estimation method with eigenvector based radial basis function neural network (RBFNN) was presented. This new method is not sensitive to noise and takes fewer training data to get RBFNN with good generalization ability. The computer simulation shows that the RBFNN trained with this algorithm is excellent at DOA estimation, and it is more suitable for real time application than classic superresolution algorithms. This DOA estimation method is promising in smart antenna of CDMA communication.