By Ben Coxworth
September 17, 2024
The prototype device identifies tiny throat movements that accompany the silent mouthing of specific words Depositphotos
There could soon be a powerful new aid for people who are unable to speak. British scientists have developed a choker that detects its wearer’s silently mouthed words, and converts them into audible synthetic speech.
Created at the University of Cambridge, the device is intended for use by people who have been rendered nonverbal due to injury, throat surgery, or conditions such as stroke, cerebral palsy or Parkinson’s disease. It could also be utilized by speaking people in noisy environments, to boost the sound of their voice without also boosting background noises.
The choker itself takes the form of a strip of bamboo-fiber fabric which is fastened around the user’s throat. It incorporates a strain sensor – located at the front – made up of two copper electrodes flanking a flexible layer of electrically conductive graphene ink. The latter is screen-printed onto the fabric.
In the current prototype, the sensor is hard-wired to a power source and computer. A commercial model could be entirely self-contained.
A closeup view of the choker’s strain sensorUniversity of Cambridge
As the user silently mouths different words, tiny vibrations in their throat cause the woven fabric to stretch and contract on a microscopic scale. As the fabric does so, an ordered network of minuscule cracks open and close in the overlying graphene layer.
That opening and closing action causes the electrical resistance of the graphene to fluctuate, which is detected in fluctuations of an electrical current flowing through the graphene from one electrode to the other.
By analyzing the telltale strength and pattern of those fluctuations, specially trained machine-learning-based algorithms on the computer are able to determine which word is being spoken. That word is then audibly voiced by a speech synthesizer.
The algorithms utilized by the choker (pictured) were trained on the speech of multiple volunteers with different accents – all of the volunteers spoke certain key words that the software learned to identify, and then used to identify other wordsUniversity of Cambridge
Thanks to the high sensitivity of the strain sensor, it’s relatively easy for the algorithms to identify each word. As a result, the system is able to maintain a speech decoding accuracy of 95.25% while reportedly using 90% less computational energy than other so-called silent speech interfaces.
“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,” says the lead scientist, Dr. Luigi Occhipinti.
A paper on the research was recently published in the journal npj Flexible Electronics.
And lest you think that the choker is the first wearable silent speech interface … well, it’s not. Past examples have included a speech-recognizing necklace, earbuds and eyeglasses.
Source: University of Cambridge
Leave a Reply