Voice personality matters
Many patients with neurological disorders, specifically neurodegenerative disorders such as motor neurone disease (MND), frequently lose the ability to communicate due to dysarthria. Not being able to speak in one's own voice leaves many patients feeling frustrated, entrapped and isolated. Current vocal communication aids are not ideal as they are often restricted to a limited set of impersonal voices that are not matched to the age or accent of each individual. Feedback from patients, carers and patient societies has indicated that there is a great unmet need for personalised voice communication aids as the provision of a personalised voice is associated with greater dignity and improved self-identity for the individual and their carers / family.
Against this background, the Centre for Speech Technology Research (CSTR) based at Edinburgh University, the Euan MacDonald Centre for MND and the Anne Rowling Regenerative Neurology Clinic started a collaborative project for voice banking and voice reconstruction. The aim of this project is to create personalised voices, for use on a communication aid, for individuals who lose the ability to speak due to a neurodegenerative disease (specifically MND). We hope that by providing a synthetic version of each individual's voice, communication and daily quality of life for people with neurodegenerative diseases can be improved.
A new technology
The current speech technology used by most vocal communication aids requires several hours of recordings just to create one voice. This is clearly a limiting factor for the creation of personalised voices for individuals with neurodegenerative diseases. The University of Edinburgh has developed a new technique of voice building that manages to retain the vocal identity of the patient with limited recordings and even deteriorating speech. This is based on two key technologies: voice cloning and voice repair.
Voice cloning allows the creation of a synthetic version of any individual's voice from a short recording of their voice (less than one hour). This technique takes advantage of the speech already collected from several voice donors so that only the essential parts of one individual's voice must be recorded. This is why one aim of this project is to record a large voice catalogue from many voice donors.
Ideally we record the patient's voice early on in the disease and create a synthetic version of his voice for use if required at a later date when speech may become affected. However, if the patient's voice has begun to deteriorate at the time of the recording, a voice repair technique can be applied in order to compensate for the disordered characteristics found in the patient's speech.