Instantaneous Bayesian regularization applied to real-time near-field acoustic holography

Area of Science:

• Acoustics
• Signal Processing
• Computational Physics

Background:

• Real-time near-field acoustic holography (RT-NAH) is crucial for analyzing non-stationary sound sources using microphone arrays.
• Traditional deconvolution methods in RT-NAH face challenges due to the ill-posed nature of inverse problems, requiring robust regularization.
• Existing regularization techniques often struggle with the dynamic and fluctuating characteristics of non-stationary sound fields.

Purpose of the Study:

• To present an instantaneous regularization process specifically tailored for the real-time near-field acoustic holography (RT-NAH) method.
• To introduce Bayesian estimation for determining the regularization parameter, leveraging prior knowledge of acoustic problems.
• To enhance RT-NAH for improved reconstruction of non-stationary sound sources and extended signal durations.

Main Methods:

• Employed singular value decomposition of the acoustic propagator for deconvolution.
• Implemented Tikhonov regularization, with a focus on instantaneous updates of the regularization parameter.
• Utilized Bayesian estimation to derive the regularization parameter based on prior information and updated it dynamically for time blocks.

Main Results:

• Demonstrated superior performance of Bayesian regularization over state-of-the-art methods in numerical and experimental reconstructions of non-stationary sources.
• Showcased enhanced RT-NAH capabilities for reconstructing longer acoustic signals.
• Confirmed that dynamic updates of the regularization parameter, linked to signal-to-noise ratio (SNR) fluctuations, are essential for highly non-stationary sources.

Conclusions:

• The proposed instantaneous Bayesian regularization significantly improves the accuracy and robustness of RT-NAH for non-stationary sound fields.
• This advanced RT-NAH technique offers enhanced capabilities for analyzing complex and time-varying acoustic phenomena.
• The method provides a more reliable approach for source identification and acoustic imaging in dynamic environments.