基于分层神经网络的宽频段DOA估计方法

该文提出了一种采用智能识别技术解决宽段来波方位估计的新方法。提取已知来波方位信号的协方差矩阵的上三角部分作为样本特征，然后构建区域估计和方位估计的分层模型，实现对未知来波方位的精确估计。所提出的分层方法降低了模型构造的复杂度，实现了宽频段来波方位估计。给出了应用该方法的具体步骤。仿真实验结果表明，该方法具有很高的测向精度，具有广阔的应用前景。

A Broadband Direction of Arrival (DOA) Estimation Approach Based on Hierarchy Neural Networks

A novel intelligent DOA estimation technique for broadband signals is proposed, which transfer the problem of DOA estimation into a large mount of data intelligent learning and recognition problem. The upper triangular half of the correlation matrix of knowing direction signals is extracted to form training set. This paper used region direction detection neural network and DOA estimation neural networks to estimate DOA for unknown signal. The main advantage of this new approach is dramatic reduction the size and complex of the training set required for each smaller neural network. The specific steps are presented. The simulation results show the DOA estimation technique for broadband signals has very high precision, and effectiveness for future application.