Media release

From: Springer Nature

Technology: A new system to produce speech *IMAGE* 

A soft patch that can transform throat muscle movements into speech with the help of machine learning is described in Nature Communications. With further development, the device could potentially help some patients with voice disorders to communicate effectively, the authors suggest.

Speech is an important aspect of human interaction, but it can be challenging for individuals with dysfunctional vocal fold conditions. About 30% of individuals in the general population experience at least one voice disorder during their lifetime. Despite this, existing solutions — such as handheld electrolarynx devices or surgery — can be inconvenient, uncomfortable, or invasive. There is a need to develop a wearable, noninvasive medical device capable of assisting patients in communicating.

Jun Chen and colleagues designed a soft magnetoelastic patch to adhere to the throat that changes shape as the throat muscles associated with speech move, without the need for functioning vocal folds. The movement of the patch detects specific muscle movements and also generates power, allowing for the device to be self-powered. These movements are then converted into electrical signals and processed by a machine learning algorithm that is trained to recognize words and translate them into speech signals. The authors demonstrated the use of this patch on eight participants without voice disorders who were asked to pronounce and whisper words and phrases, such as “Merry Christmas” or “I love you”, while standing, walking, and running. They found that the device worked with an accuracy of about 95%.

After further testing in patients, the device may offer a practical solution for individuals with voice disorders and help to enhance their overall quality of life.