A new device, created at the University of California, uses artificial intelligence to detect muscle movements in the larynx - even if, as the result of surgery, the vocal cords are no longer there - and transforms them into words.
People who suffer from voice disorders due to pathological vocal fold conditions or to laryngeal cancer surgeries often have difficulty speaking or, in the most severe cases, lose their power of speech entirely.
Now there is new hope, thanks to research conducted by bioengineers at the University of California, Los Angeles, led by Jun Chen (one of the world's most influential scientists), who have developed a sticking plaster capable of detecting movements in the muscles of the larynx—even when the vocal cords are no longer present or usable—and transforming these into electrical waves that correspond to words that people with voice problems would have trouble saying aloud. These electrical signals are then sent to an artificial intelligence (AI) program that, in turn, translates them into comprehensible sounds.
As the researchers explained in the scientific journal Nature Communications, the sticking plaster is modular, square-shaped, with sides three centimetres in length, a thickness of 1.5 millimetres, and a weight of seven grams. The two outer layers are composed of a particular elastic material (a polymer) called polydimethylsiloxane, or PDMS, and beneath each of these is an interior layer made of flexible copper electrodes. At the centre of these four lies a single layer made up of a sort of liquid containing micro-magnets, whose purpose is to generate a magnetic field.
When the person wearing the device moves the muscles in their larynx, producing only a very soft sound or none at all, the mechanical signals communicated by the muscles themselves modify the sticking plaster's magnetic field, thanks to a magnetoelastic mechanism that the University of California researchers developed in 2021 (which has already been used in a glove-like device capable of translating American Sign Language into spoken English in real time). The electrical field variations are then transformed, as described above, into electrical signals and sent to an AI program, which has been trained using information about the connections between movements of the larynx and the syllables/words these produce.
The sticking plaster is intended to be worn over the area where the vocal cords are located.
In order to test the device’s capabilities, the researchers conducted trials on eight healthy volunteers, who spoke five phrases, including “I love you,” and “Hi, Rachel, how are you doing today?” The sticking plaster detected the laryngeal movements that corresponded to these words with a very high accuracy, nearly 95%.
The researchers are augmenting the AI program with new phrases, so that it can continue to improve the scope and quality of its translations. Over the coming month, new trials will be launched involving people with language disorders.
Voice disorders, while varying in severity, are common (it is estimated that nearly 30% of people experience at least one such condition during the course of their lives). In many cases they are treatable, using a variety of therapeutic approaches, but there can sometimes be long periods of time, ranging from three months to a year—especially in cases where there has been invasive surgery—during which the vocal apparatus must be allowed complete “rest”.
“Existing solutions, Professor Chen says, like handheld electrolarynx devices, the ‘talk box’ device or tracheoesophageal puncture procedures, can be uncomfortable or invasive. This new sticking plaster is a wearable and “lightweight” device capable of assisting patients in communicating during the pre- and post-treatment recovery of voice disorders.”
The research is funded by the US National Institutes of Health.