Communication is frustrating at best for the speech-impaired. Imagine that you can't say "Good morning" or ask for a glass of water. Now imagine there's an app that can understand you and repeat what you say in intelligible speech.
Voice recognition programs – for instance, Siri – are based on the assumption that speech is standard, that everybody speaks more or less the same. They can't deal with people who stutter or have some other speech impediment.
But there's no solution out there for the speech-impaired, which led Danny Weissberg to found Voiceitt two years ago together with Stas Tiomkin, who holds a doctorate in voice recognition.
The Talkitt software they're designing should be no less than a life-changer for a host of people: "Children with autism or cerebral paralysis. Adults who suffered a stroke or have a degenerative disease such as ALS, Parkinson's, or throat cancer. And of course the deaf, some of whom cannot speak clearly," says Weissberg.
"We're a business company and want to be profitable, but we are confident we can bring heart into the business, providing a voice for people with motor or cognitive difficulties, doing something right and making a profit while about it, too," he adds.
Their personalized speech-aid system is still in the "alpha" phase of development, says Weissberg. They're taping people with speech impediments and testing their algorithm, and so far – under these "laboratory" conditions – their results have been excellent, he says.
The Talkitt technology is being tested at several institutions, says Weissberg, including Hadassah Hospital and the Sacred Family Hospital in Nazareth, which handles special-needs children with the help of funding from the Vatican.
The first stage of the product will be confined to the specific words that the user teaches it. "The problem with speech dysfunction is that the pathology can't be predicted, so our engine has been designed to be 'speaker dependent,'" explains Weissberg.
The user teaches the software a number of words – say, for sake of argument, about 20 – by saying the words and typing them in one at a time. Severely impaired people would have a caregiver around a great deal of the time who can help with the input.
Even a small bank of words can be huge for the speech-impaired, Weissberg says.
The first version of the product will only identify words that the user inputs. Weissberg forecasts that beta-stage testing with actual users can start in the first quarter of 2015, and that the product can be available for sale in the second quarter.
At a later stage, he envisions the program with learning ability: Once the software understands how the user pronounces a given phoneme (the basic unit of speech), it can start making educated guesses at all the words he uses, Weissberg explains.
Though the company is young, Weissberg reports keen interest, from people who want the product to clinicians, eager to participate in beta testing (which hasn't started yet).
Their business model, if and when, will be based on a flat fee per month for service, possibly $19, which they hope can be covered by health insurance.
The company isn't actively looking for outside capital, following Weissberg's bitter experience with shareholders in another startup, though potential investors have been calling and he has been meeting with them. He prefers to compete in "startup contests," which he's been winning. The company placed first in Partner Communications' Orange 4G Innovation Lab contest, for which it received 100,000 shekels (about $27,000). It also ranked first in a Deutsche Telekom contest, Telekom Innovation, and in Israel Mobile Summit 2014.
So far, Voiceitt has raised $500,000, including a grant from the Chief Scientist at the Economy Ministry, and is working on raising money from the general public through the Indiegogo website. So far, it's raised half its target of $40,000.
Want to enjoy 'Zen' reading - with no ads and just the article? Subscribe todaySubscribe now