Spatial Interpolation of Room Impulse Responses Using Compressed Sensing

Measuring a large set of room impulse responses inside a volume ofinterest is time-consuming unless a large number of microphones isinvolved. However, increasing the number of microphones requiresmore hardware and raises effort, e.g., in calibration. Instead of mea-suring at any desired position, it is possible to spatially interpolatethe sound field between sampled positions, in order to obtain es-timates at unknown positions. Nevertheless, the Nyquist-Shannonsampling theorem should be met, which still demands a large num-ber of spatial sampling points for large bandwidths. In this paper,we present a compressed-sensing approach that allows for stable androbust interpolation of room impulse responses using less measure-ments than required by the sampling theorem. Based on a smallset of spatially subsampled room impulse responses, the proposedmethod is capable of providing an enlarged set allowing for aliasing-free reconstruction in space.