Northeastern researcher is giving a voice back to those who lost it, literally.
By Deepak Chitnis
WASHINGTON, DC: Rupal Patel, an associate professor at Northeastern University, is behind a revolutionary new synthetic voice technology that could permanently change the way people with speech disabilities communicate.
Along with fellow research scientist Tim Bunnell, Patel has developed a way to use the sounds that people make and create a customized, synthetic voice from it, rather than a generic computerized voice. By giving people who can’t speak the ability to have their own unique and identifiable voices, Patel has revolutionized the field of computerized linguistics.
The process works by first collecting the sounds that the subject is able to make, which can range from full words to simple “oohs” and “aahs.” Each of these sounds has their own unique pitch, timbre, and so on. Patel and Bunnell then use an organization they founded called VocaliD, which collects hours upon hours of recordings from people who can speak. These records are usually readings of popular and classic books, and are used to collect samples of all the kinds of sound a human can produce.
Once both those steps are done, the sounds of the subject and the recordings of a voice that sounds similar to theirs are put through a program, also called VocaliD, which combines the two to create an entirely new voice that, although still synthetic-sounding, has some amount of personality and is readily identifiable as belonging to a single person.
The process is time-consuming, with anywhere between 800 and 3,500 sentences required to be read in order for the final voice to even stand a chance at being usable. But the donor list is already long, with Patel telling The Atlantic that some 10,000 donors have voiced (no pun intended) an interest in recording their own vocalizations for someone who can’t speak. And several hundred other people without voices have signed up to get one as soon as possible.
Patel calls the technology a “prosthetic voice;” in a TEDWomen talk Patel gave in San Francisco last year, Patel likened giving someone their own voice to giving a person the correct artificial limb.
“We wouldn’t dream of fitting a little girl with the limb of a grown man; so why, then, a prosthetic voice?” she asked.
During the talk, Patel told the touching story of a nine year-old boy she worked with named William. When the boy first tested his new voice, his first words were “Never heard me before.” Patel said that that experience – of finding out your own voice after using someone else’s for so many years, is the reason she wants to give all those with speech disabilities the means to find their own voices.
Patel is currently an associate professor in Northeastern’s Department of Speech Language Pathology and Audiology. She has a B.Sc. degree from Canada’s University of Calgary, and earned both her master’s and Ph.D. degrees from the University of Toronto. Subsequently, she was a post-doctoral research fellow at the Massachusetts Institute of Technology (MIT).
Her research interests include “prosodic control in neuromotor speech disorders, acquisition of prosody in typically developing children, design and development of human-computer interfaces, [and] assistive technology for teaching and learning language,” all of which inform the work she’s currently doing with prosthetic voices.
Patel says her goal is to have one million usable and ready-to-go voices by 2020, a large but not insurmountable goal. Working in conjunction with Bunnell, and the several donors who have already said they’re willing to participate, Patel will likely find herself giving voices to an entire generation of people who thought they’d never get the chance to hear what they sound like.
Below is the TEDWomen talk, given by Patel in December of 2013: