Abstract
Realistic sound is essential in virtual environments, such as computer games and mixed reality. Efficient and accurate numerical methods for pre-calculating acoustics have been developed over the last decade; however, pre-calculating acoustics makes handling dynamic scenes with moving sources challenging, requiring intractable memory storage. A physics-informed neural network (PINN) method in one dimension is presented, which learns a compact and efficient surrogate model with parameterized moving Gaussian sources and impedance boundaries and satisfies a system of coupled equations. The model shows relative mean errors below 2%/0.2 dB and proposes a first step in developing PINNs for realistic three-dimensional scenes.
Highlights
In computer games and mixed reality, realistic sound is essential for an immersive user experience
We take a data-free approach where only the underlying physics is included in the training and their residual minimized through the loss function, allowing insights into how well physicsinformed neural network (PINN) perform for predicting sound fields in acoustic conditions
We test the frequency-dependent impedance boundary condition, where the boundary is modeled as a porous material mounted on a rigid backing with thickness dmat 1⁄4 0:10 m with an air flow resistivity of rmat;phys 1⁄4 8000 NsmÀ4
Summary
In computer games and mixed reality, realistic sound is essential for an immersive user experience. For real-time applications spanning a broad frequency range, the IRs are calculated offline due to the computational requirements. For dynamic, interactive scenes with numerous moving sources and receivers, the computation time and storage requirement for a lookup database become intractable (in the range of gigabytes) since the IR is calculated for each source-receiver pair. Previous attempts to overcome the storage requirements of the IRs include work for lossy compression, and lately, a novel portal search method has been proposed as a drop-in solution to pre-computed IRs to adapt to flexible scenes, e.g., when doors and windows are opened and closed.. A recent technique for handling parameter parameterization and model order reduction for acceleration of numerical models is the reduced basis method (RBM).. RBM cannot meet the runtime requirements regarding computation time for virtual acoustics Previous attempts to overcome the storage requirements of the IRs include work for lossy compression, and lately, a novel portal search method has been proposed as a drop-in solution to pre-computed IRs to adapt to flexible scenes, e.g., when doors and windows are opened and closed. A recent technique for handling parameter parameterization and model order reduction for acceleration of numerical models is the reduced basis method (RBM). very efficient, RBM cannot meet the runtime requirements regarding computation time for virtual acoustics
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
