A recent study of wind noise data suggests wind noise suppression (WNS) algorithms face the largest challenge from wind noise coming from right in front of the user. As a result, the audio quality as a result is significantly worse than all other directions. This study also shows the wind noise does not contaminate all the mics equally, due to the shape of the mic array. Assuming we have some mechanism to detect the most contaminated mics and exclude them from beamforming, the output audio would be significantly improved. Therefore, we created a strategy for an audio system to handle the most challenging case with a detector for wind noise coming from the front. Upon the detection, an audio system switches to a different subset of the microphones to apply WNS, beamforming, and further noise processing.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.