homeService

A demonstrator system consisting of a microphone and a tablet-based interface has been deployed in these homes and links back to speech recognition systems running on central servers, in-the-cloud.

The homeService project is addressing the challenge of providing speech-driven control systems for older people and people with speech disorders. The robustness of these systems needs to be improved for a home environment in which there will be other noises (for instance from the TV) and a distant microphone. Speakers with disordered speech have more variable and irregular productions, and there is usually very limited training data for the speech recognition system. Each homeService system is being adapted and tailored to its user’s voice and the devices it is to control.

Data collected from people’s interaction with their systems have been collected and is available for researchers here.

The homeService project is part of the EPSRC Natural Speech Technology (NST) Programme Grant.

The network partners are:

• The University of Sheffield
• University of Edinburgh
• University of Cambridge