- Stefan Meier
In a typical cocktail party scenario it is very difficult for hearing impaired persons to concentrate on their conversation partners as the normal ability to spatially select the sources is drastically reduced. Therefore, hearing aids have to „clean up“ the microphone signals in the sense that stationary and nonstationary noise and interfering speakers are suppressed, while simultaneously preserving the intelligibility of the desired source(s).
The first attempts to improve noisy signals were all based on spectral subtraction using a single microphone. Well-known methods are the Wiener filter or statistical methods like MMSE or MAP estimators. In this thesis we want to investigate a more recent alternative approach based on eigenvalue decomposition of the estimated autocorrelation matrix for one or several microphones and apply this algorithm to monaural and binaural hearing aids. The eigenvalue decomposition (EVD) shall be investigated for speech being corrupted by nonstationary noise. First, a theoretical analysis as well as an experimental investigation by Matlab simulations is necessary. Then the performance of the EVD approach shall be compared to other well-known noise reduction techniques. Well-documented and well-structured software is important.
Matlab, Digital Signal Processing course, interest in audio
As soon as possible