New app gives throat can­cer pa­tients their voice back

Tech­nol­ogy uses pa­tient’s voice record­ings to cre­ate syn­thetic speech

Kuwait Times - - Health & Technology -

PRAGUE: Vlas­timil Gu­lar’s life took an un­wel­come turn a year ago: mi­nor surgery on his vo­cal cords re­vealed throat can­cer, which led to the loss of his lar­ynx and with it, his voice. But the 51-year-old fa­ther of four is still chat­ting away us­ing his own voice rather than the tinny tim­bre of a ro­bot, thanks to an in­no­va­tive app de­vel­oped by two Czech uni­ver­si­ties. “I find this very use­ful,” Gu­lar said, us­ing the app to type in what he wanted to say, in his own voice, via a mo­bile phone.

“I’m not very good at us­ing the voice pros­the­sis,” he added, point­ing at the hole the size of a large coin in his throat. This small sil­i­con de­vice im­planted in the throat al­lows peo­ple to speak by press­ing the hole with their fin­gers to reg­u­late air­flow through the pros­the­sis and so cre­ate sound. But Gu­lar prefers the new hi-tech voice app. It was de­vel­oped for pa­tients set to lose their voice due to a la­ryn­gec­tomy, or re­moval of the lar­ynx, a typ­i­cal pro­ce­dure for ad­vanced stages of throat can­cer.

The joint project of the Univer­sity of West Bo­hemia in Pilsen, Prague’s Charles Univer­sity and two pri­vate com­pa­nies-Cer­tiCon and SpeechTech-kicked off nearly two years ago. The tech­nol­ogy uses record­ings of a pa­tient’s voice to cre­ate syn­thetic speech that can be played on their mo­bile phones, tablets or lap­tops via the app. Ide­ally, pa­tients need to record more than 10,000 sen­tences to pro­vide sci­en­tists with enough ma­te­rial to pro­duce their syn­thetic voice. “We edit to­gether in­di­vid­ual sounds of speech so we need a lot of sen­tences,” said Jin­drich Ma­tousek, an ex­pert on text-to-speech syn­the­sis, speech mod­el­ing and acous­tics who heads the project at the Pilsen univer­sity.

A mat­ter of weeks

But there are draw­backs: pa­tients fac­ing la­ryn­gec­tomies usu­ally have lit­tle time or en­ergy to do the record­ings in the wake of a di­ag­no­sis that re­quires swift treat­ment. “It’s usu­ally a mat­ter of weeks,” said Barbora Re­pova, a doc­tor at the Mo­tol Univer­sity Hos­pi­tal, work­ing on the project for Charles Univer­sity. “The pa­tients also have to tackle is­sues like their eco­nomic sit­u­a­tion, their lives are turned up­side down, and the last thing they want to do is to make the record­ing,” she told AFP.

To ad­dress these dif­fi­cul­ties, sci­en­tists came up with a more stream­lined method for the app, which is sup­ported by the Tech­nol­ogy Agency of the Czech Repub­lic. Work­ing with fewer sen­tences-ide­ally 3,500 but as few as 300 — this method uses ad­vanced sta­tis­ti­cal models such as ar­ti­fi­cial neu­ral net­works. “You use speech models with cer­tain pa­ram­e­ters to gen­er­ate syn­the­sized speech,” said Ma­tousek.

“Hav­ing more data is still bet­ter, but you can achieve de­cent qual­ity with less data of a given voice.” The sen­tences are care­fully se­lected and in­di­vid­ual sounds have to be recorded sev­eral times as they are vpro­nounced dif­fer­ently next to dif­fer­ent sounds or at the be­gin­ning and end of a word or sen­tence, he added. So far, the Pilsen univer­sity has recorded 10 to 15 pa­tients, ac­cord­ing to Ma­tousek. Be­sides Czech, the Pilsen sci­en­tists have also cre­ated syn­the­sized speech sam­ples in English, Rus­sian and Slo­vak.

Baby di­nosaurs

Gu­lar-an up­hol­sterer who lost his job due to his hand­i­cap-man­aged to record 477 sen­tences over the three weeks be­tween his di­ag­no­sis and the op­er­a­tion. But he was stressed and less than sat­is­fied with the qual­ity of his voice. “Throat can­cer pa­tients of­ten suf­fer from some form of dys­pho­nia (hoarse­ness) be­fore the surgery, so in com­bi­na­tion with a lim­ited speech sam­ple it makes the voice sound un­nat­u­ral,” said Re­pova.

In a stu­dio at the Pilsen univer­sity mean­while, en­tre­pre­neur Jana Hut­tova is record­ing out­landish phrases. The 34-year-old mother of three faces the risk of los­ing her voice to mi­nor throat surgery-an op­er­a­tion on her parathy­roid gland. “The Chechens have al­ways pre­ferred a dag­ger-like Kalash­nikov,” she says, read­ing from the text be­fore her. “I have small kids and I want them to hear my own voice, not a ro­bot,” Hut­tova said. Then she moved on to her next sen­tence: “We were at­tacked by a tyran­nosaur’s baby di­nosaurs.”

Con­nected to the brain Ma­tousek be­lieves that in the fu­ture, pa­tients will be able to use the app to record their voice at home us­ing a spe­cial­ized web­site to guide them through the process. And he hopes that one day it will go even fur­ther. “The ul­ti­mate vi­sion is a minia­ture de­vice con­nected to the brain, to the nerves linked to speech-then pa­tients could con­trol the de­vice with their thoughts,” he said. This kind of ad­vanced so­lu­tion is a very long way off, said Re­pova.

“But look at cochlear im­plants — 40 years ago when they started, we had no idea how it would de­velop, how widely they would end up be­ing used,” she said, re­fer­ring to the in­ner-ear im­plants used to tackle se­vere deaf­ness. “A happy end would be a de­vice im­planted in the throat that could talk with the pa­tient’s own voice,” she told AFP. “It’s re­al­is­tic: it may not come in a year or even in 10 years, but it’s re­al­is­tic and we’re on the way.”—AFP

Vlas­timil Gu­lar, who lost his lar­ynx and his voice, speaks in his own voice via a mo­bile phone us­ing the spe­cial app to type in what he wants to say, dur­ing an in­ter­view at his home in the vil­lage of Mlada Voz­ice. — AFP

Newspapers in English

Newspapers from Kuwait

© PressReader. All rights reserved.