Sparse Bayesian Learning for Acoustic Source Localization

Abstract

The localization of acoustic sources is a parameter estimation problem where the parameters of interest are the direction of arrivals (DOAs). The DOA estimation problem can be formulated as a sparse parameter estimation problem and solved using compressive sensing (CS) methods. In this paper, the CS method of sparse Bayesian learning (SBL) is used to find the DOAs. We specifically use multi-frequency SBL leading to a non-convex optimization problem, which is solved using fixed-point iterations. We evaluate SBL along with traditional DOA estimation methods of conventional beamforming (CBF) and multiple signal classification (MUSIC) on various source localization tasks from the open access LOCATA dataset. The comparative study shows that SBL significantly outperforms CBF and MUSIC on all the considered tasks.

Parametric Models for Doa Trajectory Localization

“…Source directions of arrival (DOA) estimation is a crucial task in many applications such as channel estimation [1], radar [2], acoustic array processing [3], smart devices [4], and hearing aids [5]. Along with conventional beamforming (CBF) [6] and multiple signal classification (MUSIC) [7], compressive sensing based sparse reconstruction [8] and sparse Bayesian learning (SBL) [9][10][11] are some popular methods for DOA estimation. Almost all localization algorithms use block-level processing where a block consists of multiple data snapshots which are processed together to estimate the source DOA.…”

“…SBL has the unique property of automatic sparsity selection and does not require regularization 16,17 . It is a highresolution method but suffers from basis mismatch and can be computationally expensive, especially when dealing with large dictionaries or high-dimensional data [18][19][20][21][22] .…”

“…While the literature contains numerous DOA estimation algorithms, most of them process multiple snapshots together to estimate fixed DOA within a block 7,8,21,[23][24][25][26][27] .…”

Spectral Domain Sparse Representation for DOA Estimation of Signals with Large Dynamic Range

“…Oftentimes, sparse representation is applied in the data domain to the covariance matrix [4][5][6][7][8][9][10][11][12][13][14][15]. Other recent popular uses of sparse representation for improved DOA estimation include the use of Bayesian Learning [16][17][18][19][20][21][22] or co-prime and nested arrays [23][24][25][26][27][28][29][30]. Recent work focusing on the application of interpolation to decrease the off-grid effects of sparse representation has also been presented [31][32][33].…”