Three-Dimensional Sound Field Reproduction Based on Multi-Resolution Sparse Representation

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.