Equivalent source method based Near-field acoustic holography using machine learning

Abstract

The equivalent source method has been one of the most commonly used methods for sound source localization. It involves equivalent sources spread over the source plane (or region). The pressure fields from these equivalent sources are usually spherical harmonics. But, the spherical harmonic fields are derived for the Sommerfeld boundary condition with no reflection or reverberation. Data-driven methods help perform sound source localization in a reverberant environment when no prior information about the surroundings is available. The methods studied are linear regression (LR) with Adam, linear regression with L-BFGS, multi-layer perceptron (MLP) with one and two hidden layers. The simulations are conducted for two monopoles in rooms with different reverberation times and compared with one norm convex optimization (L1CVX). It is observed that overall, LR with L-BFGS gave the best results. Also, for low reverberation time, LR with L-BFGS was able to localize the sources better than L1CVX.

A combined sound field reconstruction method for non-cylindrical rotative surfaces based on statistically optimal cylindrical near-field acoustic holography and multipoint Helmholtz equation least square

“…Saijyou (Saijyou, 2004) proposed Patch NAH based on K-space and boundary element method. Statistical optimal near-field acoustic holography (SONAH), a Patch NAH method proposed by Hald (Hald, 2003), uses the linear superposition of the sound pressure or particle velocity on the holographic surfaces to reconstruct the sound. Later Li (LI et al, 2004, 2005a) studied SONAH for spherical and cylindrical surfaces.…”