Rauschverminderung zur Restauration von Audiosignalen

This master thesis Noise Reduction for the Restauration of Audio Signals deals with the restoration of audio signals by means of digital signal processing. In this context, signals from older analogue recordings (e.g. on record, tape or wax cylinder) are processed to improve the sound quality. The study begins by providing an overview of different kinds of audio signal degradations and restoration possibilities.
This thesis deals primarily with reduction of stationary broadband noise. Three well-known methods (spectral subtraction, noise reduction using subjective spectra, Wiener filter) which process this kind of signal degradation in the frequency domain, using a short-time Fourier transform, are discussed in detail. A new algorithm for broadband noise reduction is presented. It is based on the method of spectral subtraction. Test results are shown and discussed to explain the algorithm's function. Two kinds of test signals are used. Clean audio signals are degraded by synthetic noise to investigate the signal distortion caused by the process of noise reduction. Results from old analogue recordings are presented to demonstrate the performance of the proposed method.