“Can I get some Feedback, please?”

Abstract:
Speech technology traditionally treats the acoustic speech signal according to a simple transmission line model. The implicit assumption of state-of-the-art speech applications is an open-loop system that consists of a transmitter (speaker), a channel (acoustic signal) and a receiver (listener). The role of speech technology is to substitute one or more elements in this system.
It can be shown, however, that this modelling assumption cannot account for a large number of natural speech behaviours. Humans listen to their own speech output and use this auditory feedback to adjust their speech production. Speakers monitor reactions from listeners and vary their speech accordingly. These are typical behaviours that require closed-loop feedback systems. It can be argued that speech technology would benefit from either imitating these behaviours, or from exploiting the mere knowledge of their presence.
In this talk, the speaker will give an overview of several pieces of feedback-related research that he has carried out over the past years. Current work on a medical speech aid that seeks to exploit the natural feedback system will be presented, as well as suggestions and ideas for the use of feedback systems in future speech technology.

The Speaker:
Dr. Robin Hofe
Speech and Hearing Research Group
The University of Sheffield