Novel Methods of Completing Missing Samples in Audio Signals

The project goal is the investigation of novel methods for recovering missing data in audio signals. The experiences of the partners from Czech Republic and Austria will be joined to obtain new theoretical results concerning data completion in audio signals and to develop on this basis new algorithms and numerical methods. The main idea is to involve a combination of modern time-frequency representations like non-stationary Gabor transforms and methods from sparse regression that incorporate prior knowledge on the signals' structure. In both approaches, psychoacoustical properties of human hearing will be exploited.