Navigable Reconstruction of Reverberant Sound Fields Using Distributed Microphone Arrays

The reconstruction of sound fields over large domains is a challenging task that often requires a large number of measurements. To overcome this challenge, sound field reconstruction methods enable sound interpolation and extrapolation from a reduced number of measurements. In this study, we propose a method that leverages the generalisable time structure of a room impulse response. The approach accounts for the different structure of the early and late parts of the room impulse response as well as the reflection density and energy decay over time. In the early part, the strongest reflections are localised and modelled with a cluster of weighted monopoles. For the late part, kernel methods are used to reconstruct the sound field based on its statistical properties and temporal decay. The prediction ability of the method is compared to a classical plane wave reconstruction, based on experimental measurements in an auditorium. The results indicate that the approach is suited to reconstructing sound fields over large domains, achieving a fair
reconstruction across space.