一种基于神经网络的宽带相干信号波达方向估计方法

A New Method of DOA Estimates for Wideband Coherent Sources Based on Neural Network

相干信号子空间处理方法（ＣＳＭ）是一种通过构造聚焦（ｆｏｃｕｓｉｎｇ）矩阵来估计宽带相干源波达方向（ＤＯＡ）的有效方法．本文基于逆传播（ＢａｃｋＰｒｏｐａｇａｔｉｏｎ）学习算法，给出了一种用神经网络来实现聚焦过程的新颖方法．利用这种方法进行ＤＯＡ估计，可以无需预先知道源信号初始信息。

This paper presents a new method of forming focusing matrices for the use of Coherent Signal Subspace Method (CSM). The proposed focusing transformation, which is constructed based on neural network, is a kind of perfect focusing transformation, i.e. focusing without preliminary DOA estimates and iterations.