An implant to translate brain signals into speech

Home Healthcare An implant to translate brain signals into speech

It is certainly a huge step forward that the research group of the University of California, San Francisco (UCSF), has taken in giving back the opportunity to speak to those who can no longer. The system sees the integration of BCI and neural networks and the results of the study, published in the journal Nature, show how it is possible to generate sentences directly from thoughts, approaching the normal rhythm of speech, that is from 120 to 150 words per minute.

Degenerative diseases such as amyotrophic lateral sclerosis or medical conditions such as locked-in syndrome lead to loss of speech and, to date, the possibilities for the patient to communicate with the outside world are still limited.

What is locked-in syndrome?

Locked-in syndrome is a condition in which the patient is conscious and awake, but cannot move or communicate due to complete paralysis of all voluntary muscles of the body. It is the result of a brain stem stroke. 

Another important result was obtained a few months ago by a different research group, when 3 paraplegic patients were able to communicate by controlling a tablet with thought, reaching an average production of 30 characters per minute.

For more information read also Controlling a tablet with thoughts: chatting, music-streaming and much more

A two-stage process

Electrodes placements and example of signal generated by the neural network.
Credits: UCSF

“Very few of us have any real idea of what’s going on in our mouth when we speak,” Edward Chang, a neurosurgeon at UCSF and co-author of the study, said. “The brain translates those thoughts of what you want to say into movements of the vocal tract, and that’s what we want to decode.”

For more information about speech read also A Journey in the Nervous System: the speech

And that’s exactly what they did. The system used consists of two stages: a first neural network decodes the brain activity of the subject, while he is pronouncing a series of sentences aloud, in a model that simulates the movements of the vocal tract. The signals are picked up through an array of 256 electrodes positioned on the surface of the brain at the speech motor cortex. Then a second neural network translates these representations into an audio signal.

The interesting aspect is that, while the decoding of brain activity must be trained on the individual person, the translation into sounds could be generalized, according to the co-author Gopala Anumanchipalli.

This BCI has so far been tested on 5 subjects undergoing epilepsy treatment. A group of volunteers was asked to transcribe what they heard, ie the sounds reproduced by the network. Listeners heard the sentences correctly 43 percent of the time when they were given a set of 25 possible words to choose from, and 21 percent when they were given 50 words.

Even if the percentage is not high, it represents an optimal starting point for the development of the technology.

Of course, such an approach could not be used to read a person’s mind, but only to detect the words that a person would like to pronounce.

I think brain-computer interfaces will have a lot of opportunity to help people, and hopefully, to help people quickly.

~Leigh Hochberg
Neurologist at Massachusetts General Hospital in Boston


Subscribe to our newsletter

To be updated with all the latest news about Brain-Computer Interfaces, subscribe!

· By entering your e-mail address and confirming the following steps, you will accept our privacy conditions.
· Check you spam folder if you don't see the confirmation mail.
Jacopo Ciampelli
Master Degree student at the University of Bologna, my passion for medicine and engineering has led me to biomedical engineering. I've always liked writing and I'm editor-in-chief for BiomedicalCuE, area of Close-up Engineering network, too. I collaborate with Vibre and I love working with BCIs because every day there's something new to discover.


Please enter your comment!
Please enter your name here