People with speech paralysis can now talk using this intelligent spelling device

Do you remember how legendary cosmologist Stephen Hawking communicated using his special screen-equipped chair? Well, that was a brain-computer interface (BCI), a device that allows a person to communicate using their brain signals.

There are approximately 70 million people across the globe who suffer from speech-related disorders. What if there was a BCI for each one of these patients that could at least spell out words, if not speak for them? A team of researchers from the University of California, San Fransico (UCSF) works on one such groundbreaking device.

They have created a neuroprosthesis (a type of BCI device that re-establish lost functions of the nervous system) that analyzes the brain activity of a user with speech paralysis. The device then translates the brain signals into single letters and spells sentences on a screen. Reading the sentences lets anyone know what the user wants to say.

This spelling-based neuroprosthesis is better and faster

Before you understand how the current BCI works, you need to know that this is not the only type of neuroprosthesis. Some communication BCIs focus on imagined hand or body movements to figure out the message a user aims to convey.

However, the proposed BCI has some advantages over the existing ones. While explaining the same, the authors wrote in the paper, “Although implementations of this (hand-movement based neuroprosthesis) approach have exhibited promising results, decoding natural attempts to speak directly into speech or text may offer faster and more natural control over a communication BCI.”

Referring to a 2021 survey that involved 28 patients suffering from speech disorders, the researchers also claim that spelling-based neuroprosthesis is a more viable and preferred communication method than hand-movement-based neuroprosthesis.

The current device is actually an upgraded version of a BCI previously developed by the researchers. It could form sentences with a word limit of 50 words because the BCI maintained only a limited vocabulary. However, the new BCI can decode brain activity more efficiently by processing more than 1000 words.

“An initial study focused on a preliminary 50-word vocabulary. While direct word decoding with a limited vocabulary has an immediate practical benefit, expanding access to a larger vocabulary of at least 1000 words would cover over 85% of the content in natural English sentences and enable effective day-to-day use of assistive-communication technology,” said the researchers.

How does the new BIS work?

The spelling-based neuroprosthesis works by first converting signals received from the brain into phonetic alphabet letters and then spelling out the letters on the screen as complete sentences. To test their previous (50-word BCI) and current approach, the researchers chose a human subject with vocal and limb paralysis. The subject was trained to silently speak code words representing different letters that the BCI could process.

During the tests, the patient sent signals to his brain by attempting to speak the code words while being silent. He then squeezes his hand to indicate that he has spelled the desired words to BIS. The device employs language modeling and deep learning to process the signals in real time. Eventually, sentences are produced on the screen, and now anybody can read the user's response.

The error percentage in the sentences stood at only 6.13 percent, and they were produced at nearly 30 characters per minute speed. “We demonstrated that a paralyzed clinical-trial participant with anarthria (speechlessness) could control a neuroprosthesis to spell out intended messages in real-time using attempts to silently speak.”

Although more research and trials with a greater number of participants are required to further confirm the effectiveness of speech neuroprosthesis, the results of the current experiment pose the spelling-driven BIS as a promising approach for patients with speech disorders across the globe.

The study is published in the journal Nature Communications.