Robustness / performance improvement for deep learning based speech enhancement for artifacts and distortion

摘要

A method of processing an audio signal is described.This method isApply emphasis to the first component of the audio signalAnd / orThe first step to apply suppression to the second component of the audio signal to the first componentAndThe second step is to modify the output of the first step by applying a deep learning based model to the output of the first step;The first step removes the artifact and / or distortion introduced into the audio signalAndA second step for perceptually improving the first component of an audio signal.An apparatus for performing this method and a corresponding program and computer readable storage medium are also described.