This paper proposes a new method to restore music signals corrupted by impulsive noise. Since the probability density function of the music signals can be approximated by Gaussian distribution, the non-linear nature of the impulsive noises becomes prominent in the cumulant spectra of the music signals. Such feature enables the use of the higher-order spectra to detect the presence of the impulse noise. The music signals can then be restored by extracting the impulse noise.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.