Impulse Noise Removal In Speech Using Wavelets
This addresses speech quality enhancement for audio processing applications, but it is incremental as it builds on existing wavelet transform techniques.
The paper tackled impulse noise removal in speech by proposing a wavelet-based method that identifies and suppresses noise using speech's slow time-varying nature and Lipschitz regularity, resulting in significant noise reduction without degrading speech quality or introducing audible artifacts.
A new method for removing impulse noise from speech in the wavelet transform domain is proposed. The method utilizes the multiresolution property of the wavelet transform, which provides finer time resolution at the higher frequencies than the short-time Fourier transform (STFT), to effectively identify and remove impulse noise. It uses two features of speech to discriminate speech from impulse noise: one is the slow time-varying nature of speech and the other is the Lipschitz regularity of the speech components. On the basis of these features, an algorithm has been developed to identify and suppress wavelet coefficients that correspond to impulse noise. Experiment results show that the new method is able to significantly reduce impulse noise without degrading the quality of the speech signal or introducing any audible artifacts.