DeWinder: Single-Channel Wind Noise Reduction using Ultrasound Sensing

TL;DR

DeWinder introduces a multi-modal deep learning approach using ultrasound sensing to explicitly detect and reduce wind noise in single-channel outdoor speech recordings, significantly enhancing noise suppression performance.

Contribution

This work is the first to utilize ultrasound sensing as an auxiliary modality for wind noise reduction in speech enhancement.

Findings

Significant improvement over state-of-the-art noise reduction models.

Effective use of ultrasonic Doppler features for wind noise characterization.

Enhanced perceptual speech quality in outdoor environments.

Abstract

The quality of audio recordings in outdoor environments is often degraded by the presence of wind. Mitigating the impact of wind noise on the perceptual quality of single-channel speech remains a significant challenge due to its non-stationary characteristics. Prior work in noise suppression treats wind noise as a general background noise without explicit modeling of its characteristics. In this paper, we leverage ultrasound as an auxiliary modality to explicitly sense the airflow and characterize the wind noise. We propose a multi-modal deep-learning framework to fuse the ultrasonic Doppler features and speech signals for wind noise reduction. Our results show that DeWinder can significantly improve the noise reduction capabilities of state-of-the-art speech enhancement models.