Generative adversarial networks with physical sound field priors

This paper presents a deep learning-based approach for the spatio-temporal reconstruction of sound fields using Generative Adversarial Networks (GANs). The method utilises a plane wave basis and learns the underlying statistical distributions of pressure in rooms to accurately reconstruct sound fields from a limited number of measurements. The performance of the method is evaluated using two established datasets and compared to state-of-the-art methods. The results show that the model is able to achieve an improved reconstruction performance in terms of accuracy and energy retention, particularly in the high-frequency range and when extrapolating beyond the measurement region. Furthermore, the proposed method can handle a varying number of measurement positions and configurations without sacrificing performance. The results suggest that this approach provides a promising approach to sound field reconstruction using generative models that allow for a physically informed prior to acoustics problems.