反褶积
波束赋形
话筒
计算机科学
噪音(视频)
机身
麦克风阵列
算法
网格
声学
声压
工程类
人工智能
数学
电信
航空航天工程
物理
图像(数学)
几何学
标识
DOI:10.1007/978-981-15-1773-0_7
摘要
Nowadays, PhasedWang, Jiayu microphone arrays have a powerfulMa, Wei capability for acoustic source localization. The conventional beamforming constructs a dirty map of source distributions from array microphone pressure signals. Compared with conventional beamforming, deconvolution algorithms, such as DAMAS, CLEAN-SC, NNLS, FISTA and SpaRSA, can significantly improve the spatial resolution but require high computational effort. The performances of these deconvolution algorithms have been compared using simulated applications and experimental applications with simple sound source distributions. However, these comparisons are not carried out in experimental applications with complex sound source distributions. In this paper, the performances of five deconvolution algorithms (DAMAS, CLEAN-SC, NNLS, FISTA and SpaRSA) are compared in an airframe noise test, which contains very complex sound source distributions. DAMAS and CLEAN-SC achieve better spatial resolution than NNLS, FISTA and SpaRSA. DAMAS need more computational effort compared with CLEAN-SC. In addition, DAMAS can significantly reduce computational run time using compression computational grid. DAMAS with compression computational grid and CLEAN-SC are thus recommended for source localizations in experimental applications with complex sound distributions.
科研通智能强力驱动
Strongly Powered by AbleSci AI