Researchers Developed An Artificial Intelligence System That Can Transform Brain Signals Into Words

  • A new artificial intelligence technology can help people communicate who can’t speak.
  • Here's how this new technology works.

Scientists are working on a new artificial intelligence system that will transform brain signals into speech or text. It could help people to communicate who have lost their speaking ability due to stroke, or a traumatic brain injury or any other neurological disease.

Scientists from the University of California, San Francisco have been working on new technology to make communication better for people with speaking disorders. They were able to produce a system that can transform the activity of brain cells into text when a person speaks. The scientists published their research in the journal Nature Neuroscience.

They used electrodes and AI to build a machine that can turn brain signals into text. This machine tracks the activity of brain cells when a person speaks by using brain implants. An artificial intelligence algorithm works on converting the signals into numbers. Then these numbers are converted into a sequenced speech by another system which is understandable by humans. They first used the activity of neurons to predict the directions which are sent by the brain to lips, jaws, throat, and tongue for producing the words while speaking. Then a system converts these predicted movements into an artificial speech made by a computer.

The team of scientists from the University of California tested this system on four people. The people being tested were epilepsy patients. They already had brain implants for their treatment. These brain implants were used to invigilate the epileptic breakdowns of these patients. The researchers asked these patients to read out loud a series of simple sentences. The sentences were as simple as a sentence like, “Tina Turner is a pop singer”. The brain activity was decoded into speech by a computer that was monitoring the brain activity of patients while speaking. The researchers also had human transcribers with them while performing these experiments.

The new system finds it sometimes difficult to convert coherent sentences into text. For instance, the new machine translated the sentence “those musicians harmonize marvelously” into the text “the spinach was a famous singer”. According to scientists, this system is still better than any other method which we have used before for the said purpose. 97% of the text produced by the computer was found correct while there was a need to correct only 3% of the text per patient. The professional human speech transcribers have a rate of 5% recorded error. These facts mark new technology more accurate.

Your brain sends a signal from the motor cortex to lips, jaws, tongue, and throat to make a sound while you speak. The new technology only works when a person speaks the sentences aloud. But the scientists hope that they will be able to produce an upgraded version of this machine which will help people communicate with us who are not able to speak at all. This will also help patients with paralysis or locked-in syndrome to communicate. Patients are not able to verbally communicate in both of these conditions.

Photo: Rawpixel

Read next: Artificial Intelligence Allows Improving The Quality Of 19th Century Video Clip, Here's How!
Previous Post Next Post