Gaussian processes for sound field reconstruction

This study examines the use of Gaussian process (GP) regression for sound field reconstruction. GPs enable reconstruction a from limited set observations based on covariance function (a kernel) that models spatial correlation between points in field. Significantly, approach makes it possible to quantify uncertainty closed form. In this study, relation and classical methods linear is examined an acoustical perspective. Several kernels are analyzed their potential reconstruction, hierarchical Bayesian parameterization introduced, which enables construction plane wave kernel variable sparsity. The performance numerically studied compared regression. results demonstrate benefits using analysis. shows overall best performance, adequately reconstructing fundamentally different fields. appears be particularly powerful when prior knowledge would not available.