Two Czech universities have developed a voice synthesis smartphone application that allows people who lose their voice by undergoing throat cancer surgery to continue speaking with their own voice, even after removal of the organ and vocal cords.
Vlastimil Gular: Laryngeal Cancer patient |
This joint project of the
University of West Bohemia in Pilsen, Charles University in Prague and two
private companies, CertiCon and SpeechTech, was launched less than two years
ago. It was named Laryngo Voice. The technology is to use patient voice recordings
to create a synthetic one that can be taken out of the phone, tablet or laptop via
an app.
At the risk of their vocal cords,
precisely located in the larynx, Patients lose the use of their voice. A
high-tech process developed by two Czech universities, however, preserves it.
More precisely, to simulate it, thanks to vocal recordings made before the
operation, which allow to model finely the sound timbre! Thanks to a tablet or
a smartphone and a voice synthesis tool, it is possible to talk with patient's family
again. And this, with his/her "real" voice!
A common procedure in this type
of tumor is laryngectomy or extirpation of the larynx. Patients who are going
to undergo this operation, should register their voice reading as many
sentences as possible. Then, scientists use statistical models to create a
synthetic discourse that can be played on mobile phones or tablets of patients
through the app.
A Long Process
It is a long process: patients
need to record more than 10 thousand sentences to provide scientists with
enough material to produce their synthetic voice, something that makes the task
difficult because often this disease demands a quick treatment, limiting the a
patient's time to make the recordings. But researchers have found how to reduce
the number of records needed, which then goes to 3,500. A number that can even
in some cases be reduced to 300.
Advanced statistical models,
including networks of artificial neurons, then screen the recordings. Voice
synthesis specialist "Jindrich Matousek" who leads this project says
- "Speech models are used with certain parameters to generate synthetic
speech - the more data we have at the start, the better, but we can achieve
decent quality from a relatively limited material,". So far, the University
of Pilsen has registered 10 to 15 patients, according to Matousek. In addition
to the Czech Republic, Pilsen scientists have also created voice samples
synthesized in English, Russian and Slovak.
Silly phrases to drive the algorithm
However, patients requiring
ablation of the larynx usually have little time and energy to perform many
recordings after becoming aware of the diagnosis. And the words must also be
recorded several times because they are pronounced differently according to
their place in a sentence. Hence, quite ridiculous pronunciation exercises.
Gular's Story
Vlastimil Gular, who lost his job
as an upholsterer because of his health problem, A minor surgery on his vocal
cords revealed a throat cancer, which led to the loss of his larynx and, with
it, to his voice, planned three weeks later, and was able to record 477
sentences before the surgery. But he recognizes that he was stressed and rather
unhappy with the quality of his voice.
(Patients with throat cancer
often suffer from dysphonia before surgery, and its effect, with a reduced
recorded sample, produces a voice whose sound is unnatural. It's better than
nothing, though.)
Now he continues chatting with
his own voice instead of the timbre of a robot thanks to the app installed on
his phone. Gular is using the app to write what he wanted to say, in
his own voice, through a mobile phone.
A dozen laryngeal patients have
already benefited from the program. In total, a dozen laryngeal patients
registered their voice at Pilsen University. A dozen other people, in good
health, also realized.
Also, a few months ago, the
Microsoft company created an app for smartphones called Soundscape that
allows blind people to navigate independently through the city. It works with
headphones connected to the smartphone and the user can set a specific goal,
for example, a store, and during the walk hear notices about the names of the
streets, intersections and corners. In addition, you can find out from the
app which objects are around or in front of it.