Until now, people living with severe speech disorders only had a few computerized voices to choose from. Spark host Nora Young speaks with Rupal Patel, who is using speech donors to create new, personalized voices that match the individual.
If you were a little girl who couldn't speak, would you want to sound like Stephen Hawking?
Hawking, the renowned physicist who has lost the ability to speak due to a neurological disorder called amyotrophic lateral sclerosis, communicates by typing with cheek movements. The typed words are then voiced by a speech synthesizer, creating Hawking's slow monotone that has become famous.
Devices similar to Hawking's can help give voice to other people with speech impairments. Most modern devices sound more human-like than Hawking's — think of Apple's Siri or your GPS — but often the voice still doesn't suit the person using it, says Canadian speech scientist Rupal Patel.
"There's a tendency that clinicians will use an adult male voice, even for a young child."
There are also such a limited number of voices available to most speech devices that several children in the same class may have the same voice.
Patel, an associate professor of speech language pathology and audiology at Northeastern University in Boston, is changing the lives of people who rely on devices to help them speak, by creating personalized voices for them.
"They're excited to sound like them," she told Nora Young, host of CBC's Spark, in an interview that airs Saturday. "They're not looking for a perfect voice. They're looking for their voice."
- Hear the edited interview or the full episode of Spark, Sunday, Feb. 23 at 1:05 p.m. on CBC Radio One
Over five to seven years, she has developed technology and algorithms that mix the voice of a speech-impaired person with that of a healthy "speech donor."
"The combination is a sound as clear as the donor, but similar in personality to the person we want to build it for," said Patel, who is originally from Calgary.
People with a medical condition that affects their speech, such as autism or a brain injury, may only be able to utter a few vowel sounds, but Patel tries to incorporate the pitch of their voice and qualities such as breathiness and raspiness into the synthesized voice.
A donor similar in age and size will usually have a similar voice and can provide the full range of sounds used in normal speech by reading a set of sentences that include those sounds. Those can be recombined later to create whatever the person receiving the new voice wants to say.
The benefits of voice banking
Patel's challenge now is finding a wide range of voice donors.
If you're thinking about donating your voice, it may not just make a difference in someone's life, Patel suggests, but could even benefit you directly one day if you happen to lose your ability to speak owing to a misfortune such as a car accident or a stroke.
"Banking your voice is very useful," she said. "Impairments can happen at any point in life."
Since many of those who could benefit from a personalized voice are children, there is a particular need for young donors.
Getting a child to donate 3,200 utterances over three or four hours, as is done with adults, is no easy task. Patel is considering different strategies that could make it work, such as collecting the voices as part of a game or via an iPhone.
Patel's goal is to be able to create the best match possible every time so those receiving the personalized voices sound like themselves.
"It's the ability and freedom of being able to choose the kind of voice you want that's powerful here."
[Listen to a speech donor's sample recording and Spark's documentary on speech donors in the audio player links at the top-left of this page.]