Intelligibility Enhancement of Noisy Speech (IENS)

Themes: Health and Wellbeing

The objective of the project is to develop a speech enhancement system which specifically aims at improving the intelligibility of the speech signal. Although the proposed research will lead to general purpose algorithms, we focus our attention on hearing aid applications, and consequently on the algorithm constraints, e.g., with respect to algorithmic delay and complexity, imposed by this application. In contrast to most existing schemes we propose a model based approach, where the noisy signal is processed in a way that maximizes the intelligibility as predicted by an intelligibility model. In addition, the project focuses on exploiting prior knowledge in the enhancement process, of speech characteristics which are known to be important for intelligibility.