基于共同稀疏贝叶斯学习的多频等效源近场声全息方法  

Multi-frequency equivalent source near-field acoustic holography method based on common sparse Bayesian learning

作　　者：张凤敏 张小正[1] 周蓉[1] 张永斌[1] 毕传兴[1] ZHANG Fengmin;ZHANG Xiaozheng;ZHOU Rong;ZHANG Yongbin;BI Chuanxing(Institute of Noise and Vibration Research,Hefei University of Technology,Hefei 230009,China)

机构地区：[1]合肥工业大学噪声振动研究所,合肥230009

出　　处：《振动与冲击》2024年第5期260-267,共8页Journal of Vibration and Shock

基　　金：国家自然科学基金面上项目(12174082)。

摘　　要：现有基于压缩感知的等效源近场声全息方法通常采用基于单频处理的单测量向量模型进行声场重建,此模型存在噪声鲁棒性较差以及重建精度不足的问题。实际中噪声源往往具有宽频特征,同一位置处不同频率的等效源源强聚集从而呈现共同稀疏特性,若充分利用源强的共同稀疏特性,可以改善重建性能。因此,提出一种基于共同稀疏贝叶斯学习的多频等效源近场声全息方法。在该方法中,首先采用多频协同处理,构建多频等效源近场声全息模型;然后为等效源源强施加共同稀疏约束,并使用共同稀疏贝叶斯学习方法求解等效源源强。与单频等效源近场声全息方法相比,所提方法可以获得更高的重建精度和更好的噪声鲁棒性。通过单极子声源仿真和小音箱试验验证了所提方法的优越性。The existing equivalent source near-field acoustic holography methods based on compressed sensing usually use a single measurement vector model based on single-frequency processing for sound field reconstruction.This model has poorer noise robustness and insufficient reconstruction accuracy.In practice,noise sources often have broadband characteristics,and equivalent source strengths with different frequencies cluster at the same position to exhibit common sparsity.If the common sparsity of source strengths was fully utilized,reconstruction performance can be improved.Here,a multi-frequency equivalent source near-field acoustic holography method was proposed based on common sparse Bayesian learning.In this method,firstly,multi-frequency collaborative processing was used to construct a multi-frequency equivalent source near-field acoustic holography model.Then,a common sparse constraint was imposed on equivalent source strengths,and the common sparse Bayesian learning method was used to solve equivalent source strengths.It was shown that compared with the single-frequency equivalent source near-field acoustic holography method,the proposed method can obtain higher reconstruction accuracy and better noise robustness;the superiority of the proposed method is verified using monopole sound source simulation and small speaker experiments.

关 键 词：近场声全息 等效源方法 共同稀疏贝叶斯学习 多频处理 

分 类 号：TB525[理学—物理]