Direction Preserving Wind Noise Reduction of B-format Signals

Adrian Herzog, Daniele Mirabilii and Emanuël A. P. Habets

IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Canada, 2021.


Noise reduction in B-format recordings is particularly challenging as it concurrently requires to suppress undesired signals and preserve the spatial properties of the acoustic environment. In particular, wind noise poses an undesirable acoustic condition outdoors. In this work, methods to reduce wind noise while limiting the spatial distortions of the original signal are proposed based on recent works of the present authors. The main contributions are the derivation of the mentioned methods for B-format signals and the usage of the dipole-to-omnidirectional power ratio to control the trade-off between the desired-signal distortion and the noise reduction. The proposed methods are evaluated using B-format wind-noise recordings.


Example 1: Noise reduction

In this example we show the noise reduction performance of three parametric multi-channel Wiener filter (PMWF) matrices [1] in combination with the proposed desired-signal distortion/noise reduction trade-off parameter based on the dipole-to-omnidirectional power ratio.

  • Beamform-and-project PMWF (BP): single-channel estimate of the desired signal convolved with the source-to-receiver transfer functions. The noise residuals are also projected onto the desired-source direction.
  • Beamform-and-project PMWF + partial mixing (BP+PM): mix the unprocessed input and the processed output to attenuate the spatial distortions of the noise residuals.
  • Direction-preserving PMWF (DP): preserves the spatial distribution of the noise residuals at the cost of decreasing the noise reduction.

The trade-off parameter $\mu = 1 + \rho \widetilde{\textrm{PR}}$ is controlled by the signal-dependent dipole-to-omnidirectional power ratio $\widetilde{\textrm{PR}}$ and the adjustable parameter $\rho$. The case $\rho = 0$ corresponds to the multi-channel Wiener filter. The power ratio $\widetilde{\textrm{PR}}$ takes values close to 0 for plane waves and close to 1 for wind noise (assuming a spatially white noise field). Therefore, by using $\mu$, we increase the noise reduction when wind noise is predominant and reduce it when the desired plane-wave signal is present. It can also be employed to reduce any kind of spatially-white noise.

The audio files are binauralized B-format signals with 1 plane-wave source (speech), added wind noise at SNR=0dB and lower bound of the matrix filters of -20dB. The wind noise was recorded with an AMBEO VR microphone. The files were generated using the SPARTA AmbiBIN VST plugin [2].

headphones Playback with headphones is recommended. If the files fail to load (red), please refresh until all files are white.

Example 2: Spatial preservation of the noise residuals

To appreciate the difference in preserving the distribution of the noise residuals, we decreased the input SNR to -6dB and $\rho$ to 4.

headphones Playback with headphones is recommended. If the files fail to load (red), please refresh until all files are white.


[1] A. Herzog and E. Habets, "Direction-preserving Wiener matrix filtering for Ambisonic input-output systems,", In Proc. ICASSP, UK, 2019

[2] L. McCormack and A. Politis - SPARTA and COMPASS: Real-time implementations of linear and parametric spatial audio reproduction and processing methods, AES Conf. Immersive and Interactive Audio, York, UK, March 2019