We use cookies to ensure our website operates correctly and to monitor visits to our site. This helps us to improve the way our website works, ensuring that users easily find what they are looking for. To allow us to keep doing this, click 'Accept All Cookies'. Alternatively, you can personalise your cookie settings.

Accept All Cookies Personalise settings

Blogs

Brain implant turns thoughts into text, with 90% accuracy

21/09/2021

Brain Machine Interfaces (BMIs) or Brain Computer Interfaces (BCIs) have been emerging steadily over the last few decades. The technology has received considerable research funding from the military, who recognise its potential for augmenting and improving human capabilities.

A BCI is a device that enables its user to interact with a computer or machine by means of brain-activity only. BCIs can be invasive, connecting directly to the brain, or non-invasive, monitoring brain activity in order to control external software or hardware. Examples of the latter are now frequent - non-invasive commercial BCIs have been available for several years for applications that include gaming and medicine.

Now, US researchers have developed, for the first time, an invasive BCI that can decode the brain signals associated with writing letters. It can then display typed versions of these letters on a screen in real time. This new BCI will help restore communication in people affected by severe paralysis. 

The research team worked with a paralysed participant who had sensors implanted in his brain. When this individual imagined writing letters, a machine-learning algorithm recognised the patterns his brain produced. The technology was then able to translate his thoughts into text, at a speed rivalling thumb typing on a smartphone. Notably, when working at 90 characters per minute, a raw accuracy of 94.1% was attained online (self-paced), and an accuracy greater than 99% achieved offline (cue-based) after auto-correct.

Estimated time to maturity: 5 to 10 years