Speech-driven Environmental Control Systems (SPECS)

The aim is to develop a novel environmental control system (ECS) for disabled and elderly people, controlled by automatic speech recognition (ASR). This is the first product of a suite of voice-controlled health technologies and, as part of the development of the ECS, we will produce a generic software toolkit to facilitate the addition of future new applications.

Disability can make everyday tasks impossible without the assistance of an ECS. Such a system can enable users to control household devices such as the TV, lights etc. However, current ECSs can often be time-consuming and physically demanding to use. We propose to couple the functionality of ECS with ASR to overcome these limitations and increase independence for patients.

Two factors have severely limited the reliable application of ASR to ECS:

  • ECSs are used in the home where varying levels of background noise can make ASR systems unreliable.
  • ECS users often have a speech disorder such as dysarthria. Current ASR systems do not perform acceptably for this ‘abnormal’ speech thereby further disadvantaging these patients.

Members of the project team have worked on the 2nd of these challenges in the recently completed STARDUST project. STARDUST was a proof-of-concept: it demonstrated that effective voice control by severe dysarthric speakers is achievable. This was made possible by a novel methodology for developing speech recognisers tailored to individuals. We will investigate how recent advances in speech technology (e.g. speaker adaptation and noise robustness) can be harnessed to increase the performance and scope of the recognisers, e.g. handling larger vocabularies and varying environmental conditions.

This project brings together a collaborative group of academic (Barnsley Hospital, University of Sheffield) and industrial partners (Toby Churchill Ltd., XoVox Communications Ltd., and Medipex) to solve the remaining technical problems and bring this technology closer to the market.

The project is funded by the Department of Health HTD programme for three years from Feb 2006.