Multi-frequency equivalent source near-field acoustic holography method based on common sparse Bayesian learning

The existing equivalent source near-field acoustic holography methods based on compressed sensing usually use a single measurement vector model based on single-frequency processing for sound field reconstruction. This model has poorer noise robustness and insufficient reconstruction accuracy. In practice, noise sources often have broadband characteristics, and equivalent source strengths with different frequencies cluster at the same position to exhibit common sparsity. If the common sparsity of source strengths was fully utilized, reconstruction performance can be improved. Here, a multi-frequency equivalent source near-field acoustic holography method was proposed based on common sparse Bayesian learning. In this method, firstly, multi-frequency collaborative processing was used to construct a multi-frequency equivalent source near-field acoustic holography model. Then, a common sparse constraint was imposed on equivalent source strengths, and the common sparse Bayesian learning method was used to solve equivalent source strengths. It was shown that compared with the single-frequency equivalent source near-field acoustic holography method, the proposed method can obtain higher reconstruction accuracy and better noise robustness; the superiority of the proposed method is verified using monopole sound source simulation and small speaker experiments. © 2024 Chinese Vibration Engineering Society. All rights reserved.