Bayesian Framework for Room Impulse Response Reconstruction using Explicit Frequency Regularisation

Immersive spatial audio applications require a detailed understanding of sound fields in large volumes. However, data acquisition is often a restrictive task due to the large number of measurements required. To address this, we propose a Bayesian framework for Room Impulse Response reconstruction from a set of limited measurements. This approach enables the inclusion of physically-grounded models and explicit regularisation based on generalisable room properties in the frequency domain. Evaluation via experimental measurements in a historical academic auditorium demonstrates an efficient and accurate reconstruction, indicating potential applications in navigable sound field reproduction and augmented reality spatial audio.