摘要
利用PSO-BP神经网络,研究了基于矢量水听器阵列的水下声源的波达方向估计。首先对阵列协方差矩阵进行实值化和特征分解,然后将信号子空间的基作为PSO-BP神经网络的输入,并作为样本数据进行训练,以降低PSO-BP神经网络的复杂度.最后将测试样本代入PSO-BP神经网络,成功地进行了DOA估计。仿真实验表明,该方法泛化性能好,解决了输入维数过大的问题,并提高了DOA估计精度,具有较强的工程应用价值。
The PSO-BP neural network is used to study the DOA estimation of underwater sound source based on vector hydrophone array. At first, the array covariance matrix is used to real-valued and decomposed. Then, the sig- nal subspace is used as the input of the PSO-BP neural network and is trained as the sample data in order to reduce the complexity of PSO-BP neural network. In the end, the test sample is verified in PSO-BP neural network and DOA is successfully estimated. Experimental results show that the method in paper is superior to the common in generalization. The method solves the problem of input dimension that is too large, improves the estimation preci- sion, and has a strong engineering application value.
出处
《传感技术学报》
CAS
CSCD
北大核心
2016年第8期1229-1233,共5页
Chinese Journal of Sensors and Actuators
基金
国家自然科学基金项目(61275120)
