91av

Tiny throat mic can detect and broadcast silently mouthed words

A small patch worn on the throat can pick up even silently mouthed speech and broadcast it, which could help some people who are unable to speak
A throat patch that can detect silent speech and act as a speaker
This throat patch can detect silent speech and act as a speaker
Q. Yang et al

A patch worn outside the throat can detect your speech and broadcast it even when you silently mouth words. It could help workers in noisy environments or people with speech difficulties communicate.

Some throat microphones already exist, but they tend to be bulky and can only detect vibrations from quiet speech, not silently mouthed words.

To improve on this, at Tsinghua University in Beijing, China, and his colleagues have created a patch just 25 micrometres thick and about a centimetre square that is attached to the outside of the throat above the larynx with medical adhesive.

The patch is connected via wires to a microcontroller just 1 centimetre by 2 centimetres in size that can be stored in a pocket along with a button-sized battery. When the patch detects the tiny vibrations of speech, the microcontroller passes the data to an artificial intelligence model for analysis.

Sound emerges from the patch itself at volumes of up to 60 decibels, the level of a normal conversation, by using sound waves from temperature fluctuations created by an electrical input from the battery. The controller and patch can be powered by the battery for several hours.

When tested by a person speaking audibly, the device could detect the correct words with an average accuracy of above 99 per cent. Even when used by a person who had had a laryngectomy, which involves removing the voice box and thus the ability to produce audible speech, the patch was more than 90 per cent accurate.

The device could help people like firefighters and pilots be understood in noisy environments such as roadsides or cockpits and allow some people who aren’t able to speak to communicate with spoken words, says Yang.

Nature Machine Intelligence

Topics: sound