Some methods for detection and interpolation of impulsive distortions in old audio recordings

Some techniques for the detection of impulsive distortions such as FFT- and wavelet-based ones were studied. A neural network algorithm was employed to the task of detection of clicks. Some methods for the interpolation of lost samples such as nonlinear interpolation and interpolation utilizing linear predictive coding were investigated. The elaborated neural network was trained in such a way that it is performing some kind of nonlinear predictive coding of the signal, providing a valuable tool for the retrieval of samples lost inside the clicks. Some new algorithms for the interpolation of lost samples is presented.