Give the gift of voice
Since delivering a lecture at TED Women in December, Northeastern University associate professor Rupal Patel has had her work cut out for her. In conjunction with her talk, Patel launched a website called Vocalid.ai in an attempt to crowd-source the creation of personalized synthetic voices for individuals with severe speech impairment.
Patel, a speech communication expert who holds joint appointments in the Bouvé College of Health Science and the College of Computer and Information Science, has been creating similar voices for a number of years. To do so, she takes a small sample of the disordered speaker’s voice and infuses it into a library of recorded vocalizations from a healthy speaker of a similar background. But so far, the process has been much different from what she now envisions.
For one thing, the current protocol requires voice donors to visit her lab and record two to three hours of speech. Additionally, the equipment used in those recording sessions is expensive and rather high technology.
“If we’re to make a real difference in the hundreds of peoples’ lives who want their own voice, we need to streamline the process,” said Patel.
Enter The Human Voicebank Initiative. Patel’s goal is to bank 1 million voice samples by 2020. Within a couple of weeks of her TED talk, she had already collected hundreds of potential donors as well as requests from a slew of people interested in having a voice created just for them. To meet the growing demand, which is sure to increase now that the video is available to the public, Patel and her team are working to create a more user-friendly voice-banking interface.
Patel and her team hope that donors will be able to use the relatively high quality microphone on their own smartphones and tablets to record themselves. They plan to turn the process into something of a game, where donors are encouraged by in-app incentives to return again and again to do more recording. Additionally, they envision creating a learning curriculum to teach young people about the science of speech as they progress through the recording process.
“There’s a lot of work to be done but we are excited about the payoff,” said Patel. “We expect this new approach to help a lot more people than we ever could have done previously.”
People interested in donating their voices can sign up at VocaliD.ai, where they may also donate their time, expertise, and financial support.