Students speak up for those who can’t
Recorded voices help people with speech impairments
As a service organization, there wasn’t a time when the Dieruff High School Key Club wasn’t busy doing a project.
Members would organize food drives, collect toys and raise money for cancer research. Once a month, they would clean up nearby Andre Reed Park and, at Christmas, hang decorations there.
That all changed when the coronavirus pandemic hit last March and Allentown School District switched to virtual learning.
“Everything got canceled,” said Jahnia Treadwell, the club’s president.
But Treadwell wanted to find a way to continue to do good even as members held meetings on Zoom.
An internet search led her to VocaliD (vocalid.ai), a Boston company that produces custom synthetic voices for people who cannot speak because of cancer and birth, neurological and other disorders.
In recent weeks, about 10 Key Club members have begun recording their voices to donate to VocaliD’s Human Voice Bank.
“I just thought it would be
something refreshing to do, something new,” said Treadwell, a senior whose long list of school activities include being a student representative on the Allentown School Board and a member of student council.
Lots of voices needed
Some 2.5 million people in the U.S. have severe impairments to speech, with many using devices to talk, according to VocaliD founder Rupal Patel, a professor with a doctorate in speech pathology now on leave from Northeastern University in Boston.
Augmented and alternative communication — AAC — devices typically come with four preset voices.
That created a situation where Patel saw a young girl and grown man speaking to each other with the same synthetic voice in the exhibit hall at an assistive technology conference.
Patel, who had been conducting research at the Communication Analysis and Design Laboratory she founded at Northeastern, became determined to find a way to add individuality to synthetic voices.
She explained that people with speechlessness can still make sounds.
“That sound has some information about their identity,” Patel said. “The way they vocalize is unique to them.”
Her research found a way to take those vocalizations and blend them with the voice of a donor who matches the age, ethnicity and demographics of the recipient, down to regional accents. The voice could then be uploaded into the AAC device.
But to do that, Patel said, she needs recorded voices — thousands and thousands of them — with about three hours of recordings from each voice.
“You have to the find the right match voice in order to blend the voice,” said Patel, who launched VocaliD in 2014.
Easy participation
In Allentown, Treadwell said the process was fairly simple to set up, with each participant creating a dashboard on
VocaliD’s website. The Key Club secured headsets through the school for a higher quality recording.
Working through Zoom, they held their first session Feb. 17, with students meeting together then breaking out on their own to make an audition recording, which lets them know they are meeting the technical requirements.
Students are given stories — maybe science fiction or fantasy — to read over 13 recordings per session. VocaliD takes the recorded sentences and breaks them into tiny bits of sound, which are captured to create a vocal sound.
“You can choose your reading level and what books you like,” Treadwell said. “You press record. You listen back and then you submit it.”
Each session takes about 10 minutes.
“You can do snippets. You don’t have to do big passages. It really adapts to people’s schedule,” Alondra Rosario, club secretary, said.
The Dieruff students plan to hold weekly sessions where they will meet on Zoom, break out to do their recordings, then meet again as a group. They plan to have 317 recordings each by the end of the school year.
Patel was unaware of the Dieruff Key Club’s efforts but she was happy to hear about them, saying schools and organizations have held similar voice drives for VocaliD.
“Tell them thank you,” she said.
She said to be successful, the voice bank needs donors from all walks of life. ASD is an urban school district that is 85% Latino and Black.
“It’s so crucial, so crucial,” Patel said. “Without that diversity you really can’t get at preferences.”
Patel also said that without donors like the Dieruff students, VocaliD would have to charge far more than $1,499 for individualized voices.
She said the synthetic voices are partially funded with grants from the National Science Foundation and National Institutes of Health.
The company also underwrites the cost with profits from its business creating synthetic voices for companies, using paid actors. Patel emphasized that no donated voice is used for business voices.
She explained VocaliD would have to pay actors to read passages, a costly measure that would also create a situation where there were far fewer voices to run through an algorithm to find a match.
“It’s a challenge finding the right kind of voice,” she said.
Happy to help
The Dieruff students are pleased to be able to join the 25,000 people worldwide who have contributed to the Human Voice Bank.
Allentown School District is the last in the Lehigh Valley to remain all virtual, and Dieruff students haven’t been together in school in a year.
While organizations like student council continue to meet online, it hasn’t been the same.
Zahi Awad, the Key Club’s treasurer, said students hold Zoom meetings to talk about events they usually plan.
“We’re trying to think about prom, but you don’t know what is happening,” said Awad, who is also a student representative on the school board.
“It’s been hard,” Treadwell said.
Key Club members said the Human Voice Bank project is giving them a way to help others.
“Key club is very hands on. We’re very community based. The fact that Jahnia was able to find this and we are able to do this at home all together — it’s very inspiring,” Alondra said.