- Maja Taseska
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 aim at extracting the desired source signal by suppressing all noise and interference components. In order to extract a desired source signal from acquired noisy mixtures, a-priori knowledge about the target source is necessary which could be e.g., location information. Therefore, in hearing aid applications it is often assumed that the target source is located in front of the hearing aid user. However, as in reality, the target source can be located anywhere and it can also be moving, target-specific information is necessary so that the target can still be extracted if it is not located in front of the hearing aid user. This target-specific information, which could be the instantaneous pitch frequency, can be used to allow a tracking of the desired source and correspondingly, a noise reduction algorithm can be designed that allows an extraction of a desired signal from arbitrary positions.
In this thesis a recent approach to estimate the instantaneous pitch frequency in multispeaker environments [Yegnanarayana, ICASSP 2010] shall be implemented and analyzed for different conditions observed in hearing aid applications. The influence of multiple active speech sources, background babble noise, and reverberation on the proposed concept shall be analyzed and evaluated in detail. A theoretical as well as an
experimental investigation by Matlab simulations is necessary. Well-documented and
well-structured software is important.