中文摘要：

为了提高矢量水听器阵列对窄带信号的DOA估计精度，运用果蝇算法优化广义回归神经网络，通过对阵列协方差矩阵实值化，并提取信号子空间的基作为样本特征进行网络训练，构建了果蝇算法优化下的广义回归神经网络，实现了基于矢量水听器阵列的水下声源的DOA估计．仿真实验结果表明，方法泛化性能较好，能解决输入维数过大的问题，且运行时间短，DOA估计精度高，具有较强的工程应用价值．

英文摘要：

In order to improve the accuracy of DOA estimation of vector hydrophone array for narrow band signal, the array covariance matrix is real-valued and the signal subspace that train the neural network as sample features is extracted. The Generalized Regression Neural Network optimized by the Fruit Fly Optimization Algorithm is built, and the DOA estimation of underwater sound source based on the vector hydrophone array is achieved. Experimental results show that the method in paper is superior to the common in generalization and running time is short. Besides, this method solves the problem of input dimension that is too large, improves the estimation precision and has a strong engineering application value.