基于BP神经网络的正四面体阵列声源定向研究

摘要

In order to track sound source target by time delay of microphone array,a sound source orientation method based on BP neural network is proposed. This orientation system uses regular tetrahedron microphone array. A BP neural network which has two-hidden layer is designed and simulated by Matlab toolbox. The experimental results show that the network can orient the sound source target in close field or in far field, and further the Bound source tracking is carried out. It has high practicability.%针对利用正四面体麦克风阵列获取的时延值实现目标声源跟踪这个问题,提出了一种基于BP神经网络的声源定向方法.设计了一个含有双隐层的BP神经网络,使用Matlab神经网络工具箱进行仿真实验,证明可以实现远场和近场的声源定向,进而进行声源跟踪,有较高的实用性.