Automatic classification and reduction of wind noise in spectral data.

Area of Science:

• Acoustics
• Signal Processing
• Environmental Monitoring

Background:

• Outdoor acoustic recordings are frequently contaminated by wind noise, a form of atmospheric turbulence.
• This noise is particularly prevalent at lower frequencies (below a few hundred Hz) and can affect data quality even with windscreens.
• Existing methods often require direct wind speed measurements, limiting their applicability.

Purpose of the Study:

• To develop an automated method for classifying and reducing wind noise in spectral acoustic data.
• To eliminate the need for direct wind speed measurements in wind noise reduction.
• To enable efficient processing of large acoustic datasets for improved data quality.

Main Methods:

• The method identifies frequency bands with the characteristic spectral slope of wind noise.
• It utilizes uncontaminated short-timescale spectra to reconstruct a clean long-timescale spectrum.
• Classification and reduction are performed automatically on spectral data.

Main Results:

• The developed method successfully classifies and reduces wind noise in spectral data.
• Validation with field-test data confirms the effectiveness of the approach.
• The technique operates without requiring measured wind speeds.

Conclusions:

• The automated wind noise classification and reduction method is effective for improving outdoor acoustic data quality.
• This approach offers a significant advantage by not requiring wind speed measurements.
• The method is suitable for large-scale application in environmental acoustics and monitoring.