Parameter optimization of forward sound propagation models using Bayesian inference for sound field control purposes

Sound field control in outdoor concerts requires accurate estimates of the transfer functions between sources and receivers. Feed forward approaches are based on direct measurements of the transfer functions in a dense grid of points. This makes them intractable for large scale situations, showing the need of propagation models in order to characterize the sound field in such large areas.

Uncertainty in the parameters introduced in the propagation models, such as meteorological, acoustical and geometrical ones, lead to inaccurate estimates of the transfer functions and therefore to a poor performance of the sound field control strategy. In this paper we present first results of the method introduced by Heuchel et al. [1] to increase the accuracy of the predictions.

The parameters of the propagation model are optimized through auxiliary measurements and Bayesian inference.