Beamforming of phased microphone array for rotating sound source localization
Beamforming of phased microphone array for rotating sound source localization
复制标题
DOI：
10.1016/j.jsv.2019.115064
复制
发表时间：
2020-02
影响因子：
4.7
通讯作者：
Wei Ma;Huan Bao;Ce Zhang;Xun Liu
中科院分区：
文献类型：
--
作者：
Wei Ma;Huan Bao;Ce Zhang;Xun Liu
相控麦克风阵列波束形成已经成为声源定位的一种标准技术。为了在频域自由空间条件下定位旋转声源，已经研究了许多算法。两个最著名的算法是由Pannert和Maier （J. Sound Vib）提出的。， 333, 2014)和Herold and Sarradj（噪声控制工程）。J., 63, 2015)。这两种算法中使用的交叉谱矩阵之间的关系尚不清楚。本文通过提出一种计算交叉谱矩阵的新方法来研究它们之间的关系。它们之间的关系可以通过不同的插值来解释，这些插值用于计算虚拟旋转阵列麦克风的声压和真实静止麦克风的声压。前者采用傅里叶插值，后者采用线性插值。与后一种算法相比，前一种算法中的交叉谱矩阵由于具有更好的频谱重建能力，计算效率较低，声源定位精度接近相等，在高频处声源强度精度更高。此外，提出了一种基于超越方程数值求解的转向向量，作为一种计算效率更高的替代算法。
Phased microphone array beamforming has become a standard technique to localize sound sources. For localization of rotating sound sources at free-space conditions in the frequency domain, a number of algorithms have been investigated. Two most famous algorithms were proposed by Pannert and Maier (J. Sound Vib., 333, 2014) and Herold and Sarradj (Noise Control Eng. J., 63, 2015). The relationship between the cross-spectral matrixes used in these two algorithms is still unclear. This paper investigated their relationship by proposing a new approach to calculate the cross-spectral matrix. Their relationship can then be interpreted from different interpolations used to calculate the sound pressures at virtual rotating array microphones from pressures at real stationary microphones. The former uses Fourier interpolation, while the latter uses linear interpolation. Compared with the latter algorithm, the cross-spectral matrix in the former algorithm has lower computational efficiency, nearly equal sound source location precision, and better sound source strength precision at high frequencies due to its better spectrum reconstruction capability. Additionally, the steering vector based on numerically solving the transcendental equation is proposed as an alternative with higher computational efficiency to the steering vector in the former algorithm.