Improved sparse Bayesian learning direction estimation algorithm for single vector hydrophones

Aiming at the problem that complex domain sparse Bayesian learning (SBL) requires a lot of computation, a sound intensity based sparse Bayesian learning (SI-SBL) method, which combine sound intensity with SBL algorithms, is proposed to estimate the source direction-of-arrivals by single vector hydrophone. The SI-SBL algorithm converts the parameter estimation process from complex domain operations to real number domain operations by using the sound intensities as an observation. Meanwhile, noise suppression is realized by using the feature that the sound pressure channel is not related to the vibration velocity noise, which further accelerates the rate of convergence of the sparse Bayesian learning algorithm, and enables the SI-SBL algorithm to obtain higher estimation accuracy and sharp spectral peaks than SBL algorithm. The simulation results show that the single vector hydrophone SI-SBL algorithm not only performs better than the SBL algorithm, but also reduces computational complexity. The experimental results show that the SI-SBL algorithm has a 25% improvement in accuracy and an 8-fold increase in computational speed compared to SBL algorithm, verifying the effectiveness of SI-SBL algorithm.