- Position identifier: ESR9
- Host partner: UoC
Objectives
Recent speech intelligibility enhancement approaches work on clean speech and under the constraint of equal signal power before and after modification. Intelligibility gains under such a constraint are questionable, since listeners perceive a potentially-discomforting increase in loudness. Furthermore, in many situations the input speech is not itself clean, so modification algorithms are required to enhance intelligibility of distorted speech. This project investigated ways to include loudness criteria in the optimisation of signal modification algorithms, and designed new approaches for enhancing the intelligibility of distorted speech. Noise reduction algorithms were evaluated in a series of field trials having as exemplars face to face communication scenarios as well in classrooms for special listener groups.
The ESR completed the project under the supervision of Professor Yannis Stylianou and Dr. George Kafentzis, with additional supervision from Dr Mirjam Ernestus (Radboud University Nijmegen, Netherlands), Prof. Deniz Başkent (University Medical Centre Groningen, Netherlands), Dr. Cǎtǎlin Zorilǎ (Toshiba Cambridge Research Lab, UK), and Dr. Chariton Papadakis (Chania General Hospital, Crete Greece).