When an aging parent with dementia speaks in unrecognizable words, or when an autistic child communicates in his own invented language, how can caregivers, teachers or family members hope to understand what the person is trying to say?
A new Israeli company called VoiceITT fills in the gaps, giving a voice of clarity to anyone with speech difficulties.
Maybe you have a stutter that prevents you from completing a full sentence, or a neurodegenerative disease that impairs speech, or you suffered a stroke that has rendered your once eloquent prose into frustrating babble.
Using VoiceITT, your own words turn into a computerized output of what you really intended to say, thanks to a real-time intonation engine. The company’s slogan is “Capture the colors of your voice.”
Creating a personal dictionary
“A child might say ‘yum, yum, yum’ when he really means to say ‘I am hungry.’ The new caregiver comes and doesn’t know this dictionary. This is a beautiful example,” says Danny Weissberg, the cofounder and CEO of VoiceITT.
To get started, the person speaks a word, and that person or the caregiver associates it with a word on the app. In this way, the user and the calibrator create a dictionary word by word. Then the system is ready to assist the user quickly, says Weissberg.
“We don’t do speech recognition. We do pattern recognition,” he emphasizes. “We don’t relate to the recordings of hundreds of people. Instead we identify the pattern of the word that is said and if it is identical or close to that which was pre-recorded.”
And, he adds, “If the person’s situation gradually improves or degrades, our model of adaptive learning will go with them.”
VoiceITT is currently in its development phase with a working demo. Weissberg thinks they are six months away from an app ready to be downloaded to an Android device, with an iOS version planned for the future.
A voice for the 10%
“The more challenging part is getting the device to run on mobile, because there is less CPU and resources. But we don’t want to restrict the user,” Weissberg tells ISRAEL21c.
He can visualize the app one day inside a device like Google Glass and smart watches.
Though existing products for people with speech impairments rely on boards or pictures, or use eye tracking to help them select words they wish to say, “To my surprise no one has a solution like ours,” Weissberg notes.
Speech-to-text and similar tools are meant for the mass market and the 90 percent of the population that talks in a standard way, he says.
“Nobody, as far as we know, is handling non-standard speech, or the speech of small children. People with heavy accents and disabilities, including hearing disabilities, are also among this 10%.”
Version for the developing world
In the developing market, Weissberg expects the solution to be the same or similar to that intended for the Western world, but the price point will be much lower.
For now, VoiceITT is working on a subscription model of about $20 per month. Weissberg is actively pursuing ways to get America’s HMOs signed on to help waive the fees for those with a proven medical need.
The company’s developers are now working on an automatic completion technology for when the system “hears” new words.
“Because we know how a user pronounces every word, when he says something new there will be a model that will try to understand what he says,” explains Weissberg previously founded the Open University Entrepreneur Club and worked for 15 years as a team leader at Check Point and then RAD Vision, two major Israeli high-tech players. He is now working with a partner to develop a way to turn the volume of the voice into controls that can operate a computer game or other applications.
VoiceITT was founded in 2012, and is based in Ramat Gan outside of Tel Aviv. The three-person company won two development grants from Israel’s Office of the Chief Scientist.
For more information, see www.voiceitt.com.