In this paper a wind noise reduction method based on nonnegative tensor factorization (NTF) is proposed to enhance the audio quality recorded using an outdoor multichannel microphone array. The proposed method first prepares learned bases for NTF by training exemplar blocks of spectral magnitudes for a series of wind noises and audio contents. Then, the spectral magnitudes of wind noise to be reduced are estimated from the exemplar blocks. Finally, a wind noise reduction multichannel filter is constructed based on a minimum mean squared error (MMSE) criterion and applied to the multichannel noisy signal to obtain the signal with reduced wind noise. The performance of the proposed method is compared with those of conventional methods using minimum statistics (MS) and nonnegative matrix factorization (NMF) for wind noise reduction. As a result, it is shown that the proposed method provides a higher signal-to-distortion ratio (SDR), signal-to-interference ratio (SIR), and signal-to-artifact ratio (SAR) than the conventional methods under various signal-to-noise ratio (SNR) conditions.
(See document for exact affiliation information.)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
