Data-driven volumetric reconstruction in acousto-optic sensing via physics-informed 3-D Gaussian splatting

Acousto-optic sensing is a powerful approach to measuring sound at a high resolution. However, the measured value is a line integral of the sound on the optical path, which limits point-wise measurement in three-dimensional (3-D) space. To solve this problem, sound-field reconstruction methods have been proposed. Promising approaches include physical-model-based reconstruction methods, which represent a sound field as a linear combination of basis functions and determine the expansion coefficients. However, they are limited by the choice of basis functions, which means each model has a suitable sound field, making it difficult to apply a single model to all sound fields. In this talk, we introduce a data-driven approach as a more flexible method. We leveraged a 3-D Gaussian splatting (3DGS) scheme for sound-field reconstruction. 3DGS represents a 3-D scene as the sum of Gaussian kernels placed in 3-D space. In our method, the 3DGS-based volume reconstruction approach is expanded to handle sound fields and introduces a Helmholtz loss in the optimization. Our experiments have revealed that the 3DGS-based approach can be applicable to sound-field reconstruction.