基于局部保持投影和RBF神经网络的DOA估计

提出了一种有效的降维构建方法改善来波到达角（DOA）估计的性能。该方法利用局部保持投影（LPP）对DOA估计用的神经网络的训练样本进行降维，以降低神经网络的复杂度，加快神经网络的训练过程。与常用的协方差矩阵上三角特征相比，在不损失有效方位信息的基础上，可以使特征维数极大地降低。数值实验表明，基于局部保持投影和神经网络的方法具有良好的估计精度和效率，同时对噪声也有较强的适应能力，能够很好地满足波达方向估计实时性的要求。

DOA Estimation Based on RBFNN and LPP

An effective dimension reduction method is proposed to improve the performance of direction of arrival (DOA) estimation. The method applies Locality Preserving Projection (LPP) to optimize the neural network for the DOA estimation. The purpose of LPP is to reduce the training samples and the complexity of the neural network. Compared with the commonly used upper triangular half of the covariance matrix, the method can reduce the feature dimension without losing any DOA information. Simulation results indicate that the performance of the proposed method based on LPP and RBF neural network is much better than that of the traditional methods in terms of estimation precision and efficiency. Furthermore, it is not sensitive about the noise. The proposed method can satisfy the real-time requirements of the DOA estimation.