The ModelTalker project is a long-term research and development effort from the Nemours Speech Research Laboratory (SRL) intended to provide intelligible and natural-sounding personal voices for users of augmentative communication devices. SRL is one of the laboratories within the Nemours Center for Pediatric Auditory and Speech Sciences (CPASS) at the Alfred I. duPont Hospital for Children in Wilmington, Delaware. The SRL conducts research to develop, evaluate, and refine speech analysis, synthesis, and recognition technology for children and others with speech and hearing disorders.
The ModelTalker System speech synthesis system uses recorded speech (either from a prospective SGD user or from a voice donor chosen by or for the SGD user) to create a unique synthetic voice. This synthetic voice is virtually unlimited, meaning it can be used to express almost anything, including words and phrases that were not recorded. We provide the software and assistance to guide users through this voice “banking” process. Once the recording is finished, the recordings will be converted to a synthetic voice. The voice, when combined with our ModelTalker Text-to-Speech system, can then be used with a growing number of SGDs and text to speech apps for mobile devices.
One of our most exciting new areas of research is the development of personalized voices for users, especially children, who are unable to record the large amount of speech needed for a standard ModelTalker voice. Potential beneficiaries of this new research are children with Cerebral Palsy, Autism, or other conditions that make their speech difficult to understand (this is called dysarthria). These children often use Speech Generating Devices (SGDs) to communicate via commercially available (typically adult) synthetic voices. With our new technology, we will be able to create personal synthetic voices that can actually sound like the voice of the child with dysarthria. See the video clip below for more about this area of research.
Over the years, aspects of our project have benefited from support from the National Institute on Deafness and Other Communication Disorders (NIH Grant Number R42 DC006193-03), the U.S. Department of Education NIDRR grant H133E080006, NSF grants IIS 1116799 and IIS 0712821, and with continued support from Nemours Biomedical Research. We are also indebted to the developers and web hosting experts at AgoraNet, Inc., a custom software and web development company located in Newark, Delaware. And we are particularly indebted to Betsy Warren who has provided such excellent customer support over the past several years.
The ModelTalker Project Team
- H. Timothy Bunnell, PhD – principal investigator
- Jason Lilley, PhD – research scientist
- Matthew Buzzell, BS – research assistant
- Maxwell Schmid, BS – research assistant
- Bill Moyers – software development
- Derek Freer – software development
Previous team members
- John Gray – primary software engineer
- Betsy Warren – customer support
- Jim Polikoff – research assistant
- Maria Joyce – research assistant
- Jane Chandlee, PhD – Post Doc
- Katie McGrath, MSW – research assistant
- Madhavi Ratnagiri, PhD – research scientist
- Teresa Mejia, BS – research assistant
- Chris Pennington – developer
ModelTalker in the news
Here’s how we are extending ModelTalker voices for children who have never been able to speak understandably and who use communication devices.