In recent years, the application of deep learning in the field of acoustics has been actively explored. In particular, many studies have focused on modeling acoustic phenomena under primarily static acoustic conditions, such as generating room impulse responses from physical room parameters using conditional generative adversarial networks. However, in real-world environments, acoustic characteristics—such as absorption coefficients—fluctuate dynamically due to factors like human movement or changes in furniture arrangement. Approaches for adapting to such dynamic changes in acoustic conditions have not yet been sufficiently investigated. In this study, we propose a conditional deep learning approach for estimating sound fields modified by changes in acoustic characteristics. The proposed model is trained to take sound pressure distributions under certain acoustic conditions as input and to generate the corresponding sound pressure distributions for arbitrary absorption coefficients. The effectiveness of the proposed method is evaluated through simulation experiments. In future work, we aim to apply the proposed method to acoustical correction and compensation for immersive audio environments, where maintaining accurate sound reproduction under dynamic conditions is essential.
Terashita et al. (Wed,) studied this question.