.png){kind=icon}
.png){kind=icon}
Return
A Low-Complexity Sparse Bayesian Acoustic Source Localization Method Based on ℓₚ-Norm Constraint
.png){kind=icon}
PRE
AI
DOI：10.1109/TIM.2025.3648095.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Sparse Bayesian inference (SBI) has emerged as a promising approach for direction-of-arrival (DOA) estimation in acoustic signal processing due to its robust statistical framework. However, conventional SBI methods often lack flexibility due to their dependence on prior models for sparsity constraints, also struggle with high computational complexity caused by covariance matrix inversion, and additionally suffer from precision degradation due to grid mismatch. To address the aforementioned issues, this study proposes an enhanced hierarchical SBI algorithm ( $\ell _{\!p}$ -IFSBI) that integrates $\ell _{\!p}$ -norm penalty. A nonconvex $\ell _{\!p}$ -norm regularization model (with $0\lt p\lt 1$ ) is constructed to control the sparsity of the model within a hierarchical Bayesian framework. Additionally, the likelihood function is reformulated through theoretical derivation to eliminate covariance matrix inversion, thereby reducing computational complexity. Furthermore, the coati optimization algorithm (COA) is introduced to perform adaptive searching for the actual source position within the signal subspace, effectively compensating for model errors caused by grid mismatch. Simulation and real-time acoustic source localization experiments show that, at a signal-to-noise ratio (SNR) of 0 dB, the proposed $\ell _{\!p}$ -IFSBI-COA algorithm achieves a root-mean-square error (RMSE) of less than 0.3° in the DOA estimation. To facilitate further research and reproduction, the source code is available at https://github.com/Xiaob0-Zhang/lp-IFSBI
Keywords:
$\ell _{\!p}$ -norm
high precision
sound source localization
sparse Bayesian inference (SBI)
Journal
IF:
5.9
Papers:
1.9W
Citations:
5.8W