An Update on Speak:Unique

The aim

When speech is impaired people often use a communication aid, such as an iPad or eye gaze machine, to speak for them. These devices come preinstalled with synthetic voices that are very generic and unnatural sounding, leaving people to feel that they have lost not just the ability to speak, but also a part of their identity.

The aim of SpeakUnique was to develop technology that could create personalised synthetic voices for use in communication aids by individuals who have lost or will lose the ability to speak.

SpeakUnique was a collaborative research project with a multidisciplinary team of researchers at the Euan MacDonald Centre for Motor Neurone Disease Research, Anne Rowling Clinic for Regenerative Neurology and the Centre for Speech Technology Research, based at the University of Edinburgh.

Our technology

The technology we developed is unique in its ability to:

  1. Create a synthetic version of an individual’s own voice with a very short recording of the user (around 40 minutes, compared to upwards of 6 hours from traditional speech synthesis technology);
  2. Repair speech where the user already has unclear speech, e.g. slurring or slowness, due to disease;
  3. Design bespoke synthetic voices for those with no intelligible speech, based on desired characteristics such as age, gender and regional accent.

Our technology relies on accent templates that we created by blending together voice recordings of people with similar characteristics such as age and accent. During the project we recorded 1,644 donor voices from all over Scotland and other parts of the world. Thank you to everyone who donated their voice!

Our findings

In 2017, with funding from Scottish Government, we conducted a clinical pilot study. The aims of this study were to:

  • measure the value of a personalised voice to an individual and their family;
  • determine if the technology could be used outside of a professional environment;
  • gather feedback on the quality of the synthetic voices created.

The pilot project concluded that

  • personalised voices help people retain dignity and a sense of control in the face of devastating and incurable diseases;
  • high quality and rapid recordings can be undertaken in patients’ own homes and people strongly prefer recording at home;
  • people prefer a personalised voice over a generic voice;
  • personalised synthetic voices are valuable not just to the individual but also to their families.

The impact of our research

The impact our research is evident in the testimonials of our users:

  • “The SpeakUnique project has given me the chance to buck the trend and recapture an aspect of myself I thought had gone forever”
  • “My son says he would recognise it straightaway as me, if someone played the voice to him without telling him who it was.”
  • “My grandchildren said, “That’s Gramps”!”

Our research has also had an impact beyond communication aid users, including:

  • Collaboration with healthcare professionals resulted in increased awareness of voice banking and synthetic voices among Speech and Language Therapists in Scotland.
  • Public engagement activities, through the Anne Rowling Clinic and Euan MacDonald Centre, helped raise awareness of the need for communication aids and the wider work happening at the research centres.
  • Educating voice donors on the impact of neurological diseases and how they affect people, as well as Alternative and Augmentative Communication (AAC) and voice banking. This included travelling all across Scotland to record a full range of regional accents.
  • Helping to shape new healthcare legislation, as part of the National AAC advisory group. This was formalised by Scottish Government as an amendment to part 4 of the Health (Tobacco, Nicotine etc. and Care) (Scotland) Bill, which states everyone who requires a communication aid should be provided with appropriate equipment and support using it. Our clinical pilot study was funded by Scottish Government in conjunction with this bill.

Press and media coverage

SpeakUnique has been featured in a number of different media outlets. We are very grateful to all the people who have helped shed light on the work we’ve been doing. A selection of our media coverage can be found at the below links:

The One Show, 20th June 2016
The Times, 18th September 2016
iNews, 8th November 2016
BBC Breakfast, 5th February 2017
MND and 22-Year Old Me, BBC, July 2017
Huffington Post, 6th June 2018
The Scotsman, 11th June 2018
The Guardian, Seven Ways to Become a Donor, 2nd July 2018
National Museum Scotland, Communication Gallery

Academic communications

We’ve presented our research at a number of academic conferences, including the International ALS Symposium (2014, Belgium and 2013, Italy), Communication Matters (2015, UK), Interspeech Conference (2015, Germany), Home Mechanical Ventilation UK Meeting (2018, UK) and ALS International Symposium (2018, UK).

We’ve also published our research in 7 journals, the papers can be found below.
Please note some of these articles may require a subscription to read.

Reconstructing Voices within the Multiple-Average-Voice-Model framework. / Lanchantin, Pierre; Veaux, Christophe; Gales, Mark J F; King, Simon; Yamagishi, Junichi. INTERSPEECH 2015 16th Annual Conference of the International Speech Communication Association International Speech Communication Association, p. 2232- 2236.

Towards Personalized Synthesized Voices for Individuals with Vocal Disabilities: Voice Banking and Reconstruction. / Veaux, Christophe; Yamagishi, Junichi; King, Simon. SLPAT 2013, 4th Workshop on Speech and Language Processing for Assistive Technologies. ISCA, 2013. p. 107-111.

Using HMM-based speech synthesis to reconstruct the voice of individuals with degenerative speech disorders. / Veaux, Christophe; Yamagishi, Junichi; King, Simon. Proceedings of INTERSPEECH 2012 13th Annual Conference of the International Speech Communication Association. 2012. p. 967-970.

Speech synthesis technologies for individuals with vocal disabilities: Voice banking and reconstruction. / Yamagishi, Junichi; Veaux, Christophe; King, Simon; Renals, Stephen. In: Acoustical Science and Technology, Vol. 33, No. 1, 2012, p. 1-5.

Voice Banking and Voice Reconstruction for MND patients. / Veaux, Christophe; Yamagishi, Junichi; King, Simon. ASSETS 11: Proceedings of the 13th International ACM Sigaccess conference on computers and accessibility. New York : ASSOC COMPUTING MACHINERY, 2011. p. 305-306.

A Comparison of Manual and Automatic Voice Repair for Individual with Vocal Disabilities. / Veaux, Christophe; Yamagishi, Junichi; King, Simon. SLPAT 2015, 6th Workshop on Speech and Language Processing for Assistive Technologies. Association for Computational Linguistics (ACL), 2015. p. 130-133.

The voice bank corpus: Design, collection and data analysis of a large regional accent speech database. / Veaux, Christophe; Yamagishi, Junichi; King, Simon. Oriental COCOSDA held jointly with 2013 Conference on Asian Spoken Language Research and Evaluation (O-COCOSDA/CASLRE), 2013 International Conference. Institute of Electrical and Electronics Engineers (IEEE), 2013.

Thank you!

This project would not have been possible without funding and support from our key collaborators:

Euan MacDonald Centre for Motor Neurone Disease Research
University of Edinburgh Centre for Speech Technology Research
Anne Rowling Regenerative Neurology Clinic
Motor Neurone Disease Association
Medical Research Council
Anne Rowling Regenerative Neurology Clinic
Scottish Government
NHS Scotland
Innovate UK