Six degrees-of-freedom room impulse response dataset measured over a dense loudspeaker grid (6DRIR-DL)

Abstract

The need for extensive room impulse response data has become increasingly important to optimize algorithms decomposing or interpolating the audio scene via multiple spherical microphone arrays (SMAs) for virtual acoustic applications. This paper describes two datasets with a total of 134k room impulse responses (RIRs), convertible to third-order Ambisonic signals. The RIRs were measured by nine SMAs surrounding a rectangular loudspeaker grid of 392 loudspeaker positions. Both sets were measured in the same semi-cuboid room but prepared for two different reverberation times. Loudspeaker positions cover a dense grid to promote realistic benchmarking of high-resolution multi-source localization techniques using multi-perspective SMA recordings, or sound field interpolation predicting the audio scene at a variable location. Datasets are SOFA files, and source code is provided for conversion into higher-order Ambisonics (HOA) as well as source localization by MEB-ESPRIT.