使用同心多环阵提升声源定位鲁棒性*

房间混响带来的多径失真是影响声源波达方向估计精度的主要因素之一。应用于环形阵列的相干信号子空间方法可以降低相干反射声带来的不利影响。该方法对环形阵进行谐波展开,并利用环谐波域导向矢量的频率无关特性聚焦各频率下的空间相关矩阵。但单环阵列存在展开系数零点,这会导致严重的噪声放大而降低定位的鲁棒性。该文提出了一种环谐波域的最小模同心多环阵,来缓解系数零点处的噪声放大问题。设计了一套麦克风阵列系统,用于评估同心多环阵列定位的鲁棒性。仿真和实验结果均表明:与相同孔径和阵元数的单环阵相比,使用最小模准则设计的同心多环阵列可以显著提升混响场景下的声源定位的稳健性。

Robust sound source localization using concentric circular arrays

One of the factors to decrease the accuracy of the direction of arrival (DOA) of the sound source is the multipath distortion caused by reverberation. The Coherent Signal Subspace Method (CSSM) applied to circular arrays can reduce the adverse effects of reflected sound. The method performs harmonic expansion on the circular array and focuses the spatial correlation matrix (SCM) at each frequency by using the frequency-independent characteristics of steering vector in the circular harmonic domain (CH). However, the circular array has nulls of expansion coefficient (nulls of Bessel function), which causes severe noise amplification and reduces the robustness of localization. In this paper, a minimum modulus concentric circular array is proposed to alleviate the noise amplification problem at the zeros of the Bessel function. A microphone array hardware was designed to verify the robustness of localization with the uniform concentric circular array. Both simulated and experimental results indicate that the concentric circular array with minimum modulus can improve the robustness of DOA when compared to the circular array with the same element number and array aperture.