AI-powered tech could help people with speech impairments to work remotely

  • CNN
  • September 23, 2024
CNN

 — 

You've probably experienced the frustration of being misheard or misunderstood by a smart speaker or AI assistant. For people with non-standard speech, it can happen in nearly every interaction with this kind of technology, Israeli company Voiceitt aims to change that.

By using personalized voice models, its AI-powered speech recognition system helps people with speech impairments, caused by conditions like cerebral palsy, Parkinson's, Down Syndrome or stroke, communicate more effectively with both people and digital devices.

For Voiceitt co-founder Sara Smolley, facilitating speech recognition for non-standard speech patterns is a personal mission.

"My grandmother was diagnosed with early-onset Parkinson's disease," she said. "By the time I was born, she had lost most of her motor capabilities, and her speech was impacted."

Voiceitt was launched as an app in 2021 and operated as a simple vocal translator, converting non-standard speech to audio. The AI is trained by the user recording themselves saying around 200 simple stock phrases.

Smolley said the original idea was to facilitate in-person communication, but the technology has now also been tailored to remote workers.

Voiceitt has developed integrations with WebEx and ChatGPT, along with a Google Chrome extension, which convert non-standard speech to captions shown on the screen. The company is also partnering with Zoom and Microsoft Teams.

"One of the things that really stood out to me was the importance of accessibility technology in the workplace," Smolley noted. "What a (wheelchair) ramp was to yesterday's office building Voiceitt is to today's remote workplace," she said.

The software is sold as either a per-minute or per-user license, with prices ranging from $20 to $50. Licenses can be bought in bulk for workplaces and health or education institutions.

"People are using Voiceitt not just for video meetings, but for writing documents, emails, posting on LinkedIn and accessing web browsers by voice," Smolley explained. "This has opened up the digital world for individuals who previously might not have been considered for certain jobs or able to communicate with colleagues or customers."

"This entire sector needs a shake-up"

Among users of Voiceitt is Colin Hughes, a former BBC producer turned accessibility advocate. Living with muscular dystrophy, Hughes relies on dictation for his digital interactions, making him acutely aware of both the potential and limitations of current technologies. Hughes has used Voiceitt to compose emails and dictate longer written pieces.

"I found Voiceitt's app to be impressively accurate with my atypical speech, and its training and setup process was straightforward," Hughes said, though he highlighted critical gaps for professional users. "Many people with impaired speech and upper-limb disabilities need more than just speech-to-text," he noted.

He advocates for features like voice-driven cursor control and improved dictation recognition for drafting long-form content, saying that Voiceitt works best for single-sentence messages.

Hughes stresses a need for more comprehensive speech-recognition tech that lets users do things like manage emails and format documents using their voice.

He sees a future where technology plays a bigger role in accessibility, adding: "This entire sector needs a shake-up. Voiceitt, with better access to major platforms, could be the one to lead this change."

An accessible future

According to Smolley, there has been significant progress in speech recognition technology in recent years.

One example is the Speech Accessibility Project, a research project led by the University of Illinois' Beckman Institute for Advanced Science and Technology, which is collecting voice data from people with different conditions to build algorithms for supporting people with non-standard speech.

Earlier this year, Apple (AAPL) launched its AI-powered "Listen for Atypical Speech" feature, which uses machine learning to recognize a wider range of speech patterns.

With technology that captures and stores personal data such as voice recordings, privacy can be a concern for users. Smolley says her company complies with European Union regulations, which she calls "the highest standards in the world in terms of data privacy."

"If the user's data is being kept in our database with their consent, it's anonymized and de-identified, and used to augment our data stack and improve our algorithm," she adds.

She believes Voiceitt's technology can be life-changing. "We want to allow people not just to be more independent in their lives and work," says Smolley, "but also to enjoy technology and have fun."