基于Fourier变换-源强模拟技术近场声全息正则化方法比较

Comparison of regularization methods of nearfield acoustic holography based on fourier transform-source simulation technique

近场声全息中通常需要采取正则化方法来改善重建的不适定性.研究在基于Fourier变换-源强模拟技术的近场声全息理论基础上,分别对截断奇异值法和Tikhonov法进行了比较,并使用广义交叉验证法和L-曲线法对正则化参数进行选取,最后采用不同的组合方法对声场重建进行仿真分析.结果表明:运用L-曲线法结合Tikhonov正则化方法,重建声场的相对误差较小,能有效地解决该理论重建过程中的不适定性.

In order to solve the ill-posed nature in the Near-field Acoustic Holography （NAH） reconstruction ,the regularizatiou procedure was usually used. In this paper, upon the Near-field Acoustic Holography based on Fourier Transform-Source Simulation Technique, a comparison was made between truncated singular value decomposition （TSVD） method and Tikhonov regularization. In addition, the regularization parameter was chosen by the L-curve criterion and the generalized cross validation （GCV） method. Finally, some acoustic reconstructing numerical simulations were taken by using different regularization methods. The results show that the sound field reconstruction relative error is smaller when using L-curve method with Tikhonov regularization method, which can effectively solve the reconstruction process of the ill-posed.