基于稀疏贝叶斯学习的高分辨率Patch近场声全息  

Super resolution patch near-field acoustic holography via sparse Bayesian learning

作　　者：扈宇 胡定玉[1] 方宇[1] 肖悦[2] HU Yu;HU Dingyu;FANG Yu;XIAO Yue(School of Urban Rail Transportation,Shanghai University of Engineering Science,Shanghai 201620,China;Jiangxi Province Key Laboratory of Precision Drive and Control,Nanchang Institute of Technology,Nanchang 330099,China)

机构地区：[1]上海工程技术大学城市轨道交通学院,上海201620 [2]南昌工程学院江西省精密驱动与控制重点实验室,南昌330099

出　　处：《振动与冲击》2018年第16期104-110,153,共8页Journal of Vibration and Shock

基　　金：国家自然科学基金(51605274;51565037);上海高校青年教师培养资助计划专项基金(ZZGCD15115);上海工程技术大学研究生科研创新项目(16KY1001)

摘　　要：针对基于空间二维Fourier变换的近场声全息技术的重建效果受到测量孔径与测点数影响的问题,提出一种基于稀疏贝叶斯学习的高分辨率Patch近场声全息方法。利用高斯核函数和稀疏贝叶斯学习算法对全息面声压进行插值和外推,利用插值和外推后的全息面声压进行近场声全息重建。仿真和实验结果表明,所提方法可有效抑制小全息孔径测量对重建精度的影响,同时可以在测点较少的情况下极大提升全息重建的空间分辨率。另外,仿真结果还证明了插值过程具有较好的稳定性,可以在一定程度上提高测量数据的信噪比。An approach for super resolution patch near-field acoustic holography was proposed based on sparse Bayesian learning.The interpolation and extrapolation models were first established by use of the Gaussian kernel functions and the sparse Bayesian learning,and then the measured pressure was simultaneously interpolated and extrapolated to obtain a larger and denser virtual measurement.Finally,the interpolated and extrapolated pressures were used to perform near-field acoustic holography.Results of the simulation and experiment show that the aperture effect was greatly suppressed and the super resolution reconstruction can be achieved when using the Fourier-based near-field acoustic holography.It also shows that the measurement noise was suppressed in the process of interpolation.

关 键 词：Patch近场声全息 稀疏贝叶斯学习 数据插值 数据外推 

分 类 号：TB532[理学—物理]