Interpolation of the early part of the acoustic transfer functions using block sparse models.

Two types of block sparse models are presented to interpolate the early part of the acoustic transfer functions (eATFs) between the loudspeakers and the target spatial regions in reverberant rooms. The desired eATFs are modeled as a superposition of sound fields generated by a dense grid of equivalent point sources, which exhibit two kinds of block sparse structures under the far-field and small target region assumptions. Simulation results in the two-dimensional case demonstrate that the proposed models can provide accurate eATF interpolation over a broad frequency range using a small number of microphones.