Researchers at the University of Cambridge have developed a 'smart choker' that could transform the lives of people with speech impairments.
Why it matters
Current speech assistance technology is bulky and rigid, sometimes requires surgery, and often requires extensive user training. Because of its AI capabilities, the Cambridge choker removes these issues with a comfortable, non-invasive design and reduced training needs.
- Help for a wide range of people, from those with temporary speech impairments due to laryngeal surgery to permanent conditions like Parkinson's, stroke, or cerebral palsy.
How it works
The choker is a silent-speech interface designed to analyze non-vocal signals and decode speech in silent conditions. It's made from a soft, stretchable, and comfortable fabric with embedded electronic sensors.
- As users mouth the words they want to speak, the device captures the subtle vibrations of the throat produced during whispering or silent mouthing of words.
- An AI model then interprets this information to create the speech intended by the user.
- The captured speech signals can then be transferred to a computer or speaker to facilitate conversation
The choker is made from a bamboo-based fabric.
The choker's performance
- The smart choker has been tested and shown to recognize words with over 95% accuracy, significantly improving existing technologies.
- It uses 90% less computational energy, making it more efficient and sustainable.
- The embedded sensors, made from graphene ink, are super sensitive and ideal for speech detection.
“These sensors can detect tiny vibrations, such as those formed in the throat when whispering or even silently mouthing words, which makes them ideal for speech detection. By combining the ultra-high sensitivity of the sensors with highly efficient machine learning, we’ve come up with a device we think could help a lot of people who struggle with their speech.” —Luigi Occhipinti, PhD, Cambridge Graphene Centre
Situational awareness
- To ensure the choker adapts to different users, the researchers trained their AI model on a database of frequently used English words, including often confused ones (e.g., book and look).
- They also trained the model with various users to account for gender, native language, accent, and speaking speed.
The takeaway
As with all new technologies, substantial testing and clinical trials are needed before the choker is approved for patient use. Nevertheless, the results offer a glimpse into a future when communication barriers might be eliminated for many.
“I was surprised at just how sensitive the device is. We couldn’t capture all the signals and complexity of human speech before, but now that we can, it unlocks a whole new set of potential applications.” —Dr. Occhipinti
