中文摘要：

研究声源定位优化建模问题,针对声源位于远场环境下无法获取精确的方位角和俯仰角,由于采用声达时间差（TDOA）和空间几何算法的正四面体麦克风阵列声源定位方法只适应于近场声源定位,为了提高定位准确性,提出了应用径向基（RBF）神经网络建立声源定位模型的算法,声源定位模型在声源位于近场或者远场的情况下,均可求解出精确的方位角和俯仰角。在MATLAB上进行仿真,结果表明,定位声源的方位角误差小于3°,俯仰角误差小于4°,满足实际定位精度的要求。结果表明为声源准确定位提供了科学依据。

英文摘要：

Study on optimization of sound source localization model. Due to tetrahedral microphone array sound source localization model based on time difference of arrival （TDOA） and spatial geometry algorithms only adapt to the far - field localization, the localization model cannot get accurate azimuth and elevation angle when the sound source is located near - field. In order to improve the localization accuracy, a sound source localization model was proposed based on radial basis function （RBF） neural network. The model can find a solution of both accurate azi- muth and elevation angle in spite of in near - field or far - field. Simulated by MATLAB, the results show that the azimuth error of sound source is less than 3 °, the elevation angle error is less than 4 °, which can meet the actual requirement of localization accuracy.