On the suppression of noise from a fast moving acoustic source using multimodality

Abstract

The problem of removing non stationary noise coming from a moving acoustic source in outdoor environment is investigated in this paper. By making use of the known instantaneous location of the moving source (provided by a second modality), we propose a time-domain method for removing the noise coming from this source in a mixture of acoustic sources. The proposed method consists in an irregular resampling of the mixed data recorded by the sensors, and by linearly combining the resampled data to remove the undesired source. It’s similar to the beamforming approach, but instead of seeking to enhance a target source, we seek to remove an undesired one. Simulation on synthetic data show the effectiveness of the proposed method with a few number of sensors.