基于改进MUSIC的声源定位方法

为设计更有效的声源定位方法,深入研究了传统 MUSIC 算法,并针对其分辨率低且在麦克风数目较少时波达方向 (direction of arrival,DOA)估计精度较差等问题,对传统 MUSIC 算法进行了优化,利用广义互相关算法估计出声源信号到达各 麦克风之间的时间差,并据此构建出对应的矢量信号,最后通过计算谱函数确定 DOA 估计值。 仿真和实验结果表明,优化后的 MUSIC 算法可以得到更加尖锐的指向性波束,更低的旁瓣,并且能使方位角的定位精度达到±4°,俯仰角的定位精度达到±5°。

Sound source localization method based on improved MUSIC

In order to solve the problem of low resolution and poor estimation accuracy of direction of arrival (DOA) when the number of microphones is small, so that designing an effective sound source localization system, this paper deeply studies and optimizes the MUSIC algorithm. The generalized cross-correlation is used to estimate the time difference between the sound source signals reaching each microphone, and building the corresponding vector signal. Finally, the DOA estimation value is determined by calculating the spectral function. The simulation and experimental results show that the optimized MUSIC algorithm can get sharper directional beam, lower sidelobe, and the positioning accuracy of azimuth angle can reach ±4° and elevation angle can reach ±5° respectively.