Three-dimensional sound field reproduction based on multi-resolution sparse representation
Three-dimensional sound field reproduction based on multi-resolution sparse representation
复制标题
基于多分辨率稀疏表示的三维声场再现
DOI：
10.1109/icspcc.2015.7338801
复制
发表时间：
2015
期刊：
影响因子：
--
通讯作者：
Kai Liu
中科院分区：
文献类型：
--
作者：
Guangfei Zhang;Qinghua Huang;Kai Liu
本文提出了一种多分辨率稀疏表示方法来再现三维高阶立体混响声场。从低到高的空间角度分辨率构造多分辨率字典。基于这些字典，虚拟平面波源的位置估计的复杂的最小绝对收缩和选择算子（CLASSO）算法。该方法能有效地定位原始信号源，并降低了计算复杂度。最后，通过去偏置步骤来调整虚拟平面波的信号以获得适当的扬声器权重。仿真结果表明，该方法可以提高3D再现性能与几个有源扬声器和扩大的空间甜蜜点的大小。
In this paper, a multi-resolution sparse representation method is proposed to reproduce three-dimensional (3D) Higher Order Ambisonics (HOA) sound fields. The multi-resolution dictionaries are constructed from low to high spatial angle resolution. Based on these dictionaries, the positions of virtual plane wave sources are estimated by complex least absolute shrinkage and selection operator (CLASSO) algorithm. The proposed method can effectively localize primary sources and reduce the computation complexity. Finally the signals of virtual plane waves are adjusted by a debiasing step to derive proper loudspeaker weights. Simulation results show that the proposed method can improve the 3D reproduction performance with a few active loudspeakers and enlarge the size of the spatial sweet spot.