Physicist Stephen Hawking has long dealt with a degenerative disease that leaves him unable to speak. Instead, he uses a speech synthesizer that bellows out his words in a now-recognizable robotic male voice. It works for Hawking–he’s an adult male. But similar robotic voices are used for little girls and all the other 2.5 million Americans unable to talk. These people would speak in dramatically different vocal registers if they had the ability.
At this month’s TEDWomen event in San Francisco, speech scientist Rupal Patel unveiledVocaliD, a project that will create prosthetic voices that actually sound like the people who use them. It could be a life-changing development in the robo-voice world.
To create a prosthetic voice, Patel and her team at Northeastern University gather as many acoustical properties as they can from the person being fitted with the voice (that is, all the sound they’re still able to produce) and overlay those properties on a voice from a surrogate voice donor who is the same gender, size, and age. A voice donor typically spends three to four hours reciting phrases that cover all the combinations of sound present in a language.