Contribution
Reconstruction of reverberant sound fields exploiting frequency autocorrelation functions
* Presenting author
Abstract:
Sound field reconstruction techniques are at the core of several applications, including navigable audio, sound field visualisation and room acoustics. However, the lack of experimental data, particularly at high frequencies and for large spatial domains, highlights the need for additional prior knowledge about the sound field structure. In this paper, we explore the use of the covariance function across frequency bins, which presents a known closed-form solution above the Schroeder frequency and depends on room parameters such as the reverberation time and the modal density. We incorporate this covariance function as prior knowledge in a Bayesian inference framework and evaluate its performance against classical approaches where frequency independence assumed. Experimental tests are conducted in a historical auditorium in Copenhagen, where the results demonstrate the relevance of the frequency covariance in probabilistic approaches.