基于联合方向特征的智能来波方向估计

Intelligent Direction of Arrivals Estimation Based on Unite-direction-character

针对利用智能学习方法进行多信号源二维方向估计模型难以构建的难题,提出了一种有效的降维构建方法。该方法首先对构造的DOA矩阵进行特征分解,获取各个信号源对应的、包含两个方向信息的特征值和特征向量,分别以这两个量(统称为联合方向特征)作为输入特征,以单信号样本来训练两个RBF神经网络模型,然后利用训练好的模型分别对分离出来的各个信号对应的特征值和特征向量进行映射估计来得到两个空间角。仿真结果表明:该方法达到了对多信号源二维来波方向进行降维估计的目的,且具备较高的估计精度。

To solve the problem of model-building for 2D-DOA estimation of multi-sources based on smart learning method, an effective dimension-degraded model-building approach was proposed in this paper. This ap- proach based on the DOA matrix decomposing could get the eigenvalue and steering-vector of each signal source, then separately took them as the input to train two radial basis function neural network （RBFNN） models using the training sets of single source, the trained models could approximate the nonlinear mapping from the separating characters of multi-sources to direction-of-arrivals, these two models could get both two angles of the arrivals. The simulation results showed it had a high estimation precision, so this method has a bright application foreground.