Study of GSO-GL based secondary source configuration optimization method for broadband sound field reproduction

宽带声场再现中基于GSO-GL的次级声源布局优化方法研究

¹ School of Marine Science and Technology, Northwestern Polytechnical University, Xi’an 710072, China
² Zhijiang Lab, Hangzhou 311121, China

Received: 16 March 2023

Abstract

Sound field reproduction is a method that uses multiple secondary sound sources to reproduce a predefined desired sound field in a listening area. The configuration of the secondary sound sources is a key factor affecting the reproduction performances. To optimize the secondary source configuration, this study proposes a new method that combines the Gram-Schmidt orthogonalization method with the group Lasso method, which is called GSO-GL method. Firstly, it selects a certain number of the secondary sources from all the alternative secondary sources using the Gram-Schmidt orthogonalization method. The excitation sources are then further selected from the secondary sources selected in the previous step using the group Lasso method. The sound field simulation from the 2D room model show that the GSO-GL method outperforms the Gram-Schmidt orthogonalization method or the group Lasso method alone, combining the high system stability of the Gram-Schmidt orthogonalization method with the high reproduction accuracy of the group Lasso method.

摘要

声场再现是一种利用多个次级声源在听音区域内再现预定义期望声场的方法, 其中次级声源布局是影响再现性能的关键因素。为解决次级声源布局优化问题, 提出一种将Gram-Schmidt正交化方法与组Lasso方法相结合的方法(称为GSO-GL方法), 即先使用Gram-Schmidt正交化方法从所有备选次级声源中选出一定数量的次级声源, 再利用组Lasso方法从前一步所选的次级声源中进一步选择出激励声源。二维房间模型仿真结果表明, GSO-GL方法的性能优于单独使用Gram-Schmidt正交化方法或组Lasso方法, 综合了Gram-Schmidt正交化方法的高系统稳定性和组Lasso方法的高再现准确度。