[Paper Review] Beamforming and other methods for denoising microphone array data

Pieter Sijtsma, Alice Dinsenmeyer+2 more|arXiv (Cornell University)|Jun 7, 2019

Aerodynamics and Acoustics in Jet Flows11 references4 citations

TL;DR

This paper evaluates beamforming and denoising techniques—Conventional Beamforming, Source Power Integration, and CLEAN-SC—for suppressing turbulent boundary layer (TBL) noise in microphone array data. By exploiting the diagonal concentration of TBL noise in the cross-spectral matrix, the study demonstrates that CLEAN-SC outperforms others in resolving weak sources under high noise conditions using synthesized array data.

ABSTRACT

Measured acoustic data can be contaminated by noise. This typically happens when microphones are mounted in a wind tunnel wall or on the fuselage of an aircraft, where hydrodynamic pressure fluctuations of the Turbulent Boundary Layer (TBL) can mask the acoustic pressures of interest. For measurements done with an array of microphones, methods exist for denoising the acoustic data. Use is made of the fact that the noise is usually concentrated in the diagonal of the Cross-Spectral Matrix, because of the short spatial coherence of TBL noise. This paper reviews several existing denoising methods and considers the use of Conventional Beamforming, Source Power Integration and CLEAN-SC for this purpose. A comparison between the methods is made using synthesized array data.

Motivation & Objective

• To address the challenge of turbulent boundary layer (TBL) noise corrupting acoustic measurements in wind tunnel and flight testing environments.
• To evaluate the effectiveness of beamforming and denoising techniques in recovering weak acoustic sources masked by spatially coherent TBL noise.
• To compare Conventional Beamforming, Source Power Integration, and CLEAN-SC in terms of source localization accuracy and noise suppression.
• To provide a performance benchmark using synthesized array data under controlled conditions.

Proposed method

• Utilizes the cross-spectral matrix (CSM) of microphone array data to model noise characteristics, with TBL noise concentrated along the diagonal due to short spatial coherence.
• Applies Conventional Beamforming to form beamformed maps and localize sources based on beam response.
• Employs Source Power Integration to estimate source power by integrating beamformed energy across frequencies and spatial locations.
• Implements CLEAN-SC, an iterative algorithm that identifies and subtracts the strongest noise components from the CSM to enhance source resolution.
• Uses synthesized array data with known source locations and TBL noise levels to simulate realistic measurement conditions.
• Compares methods using metrics such as source localization accuracy, dynamic range, and signal-to-noise ratio improvement.

Experimental results

Research questions

• RQ1 How do Conventional Beamforming, Source Power Integration, and CLEAN-SC perform in resolving weak acoustic sources under high TBL noise?
• RQ2 To what extent does the diagonal concentration of TBL noise in the cross-spectral matrix enable effective denoising?
• RQ3 Which method provides the best trade-off between source localization accuracy and noise suppression in simulated microphone array data?
• RQ4 How does CLEAN-SC compare to beamforming-based methods in resolving closely spaced or low-amplitude sources?

Key findings

• CLEAN-SC significantly outperforms Conventional Beamforming and Source Power Integration in resolving weak sources under high TBL noise conditions.
• The diagonal concentration of TBL noise in the cross-spectral matrix enables effective noise suppression, particularly when using iterative methods like CLEAN-SC.
• Source Power Integration provides moderate noise suppression but struggles with source resolution when noise levels are high.
• Conventional Beamforming is effective for strong sources but fails to resolve weak or closely spaced sources in the presence of TBL noise.
• Synthesized data results confirm that CLEAN-SC improves the dynamic range of source maps and reduces false detections.