The interface is a record holder
The new implant allowed the paralyzed person to write text at almost normal speed
Sergey Vasiliev, Naked Science
Neurophysiologists have managed to turn the brain's mental signals into written text in real time. The new implant and computer system work more than twice as fast as analogues, allowing you to record up to 90 characters per minute. Scientists write about this in an article published in the journal Nature (Willett et al., High-performance brain-to-text communication via handwriting).
The novelty was created by the BrainGate research consortium, whose participants develop various types of brain–computer interfaces and recently tested one of them at home. The project was led by Jaimie Henderson from Stanford University. A new system has also been tested in the laboratory. By reading and interpreting signals from the motor cortex, which normally controls hand movements, she allowed the paralyzed patient to record about 16 words per minute.
The algorithm recodes the squiggles "drawn" by the patient into easy-to-read characters. You can see the final result in a video on YouTube – VM.
The volunteer was a 65-year-old man with spinal cord injury. Two miniature sensors measuring only 4 x 4 millimeters were fixed on the surface of his cerebral cortex. Each of them carried an array of 100 thinnest electrodes that were immersed in the motor cortex.
Array of electrodes used in the work.
Accordingly, each of the pair of sensors registered signals from about a hundred neurons, sending them to a computer.
The program was trained to recognize patterns of cell activity at moments when the patient imagined that he was writing a text by hand. By agreement, the space between the words was indicated by the written symbol ">". The system prepared in this way was able to recognize letters with an accuracy of about 95 percent at a speed of 16 words per minute. According to scientists, this is about 3/4 of the normal typing speed of people over 65 years old.
Unfortunately, even such a highly efficient interface can only be called a concept. Installing sensors requires a dangerous invasive procedure, besides, the system has to be pre-trained for each patient individually, and its neural networks need powerful and not cheap computing resources. However, it is the creation of such brain–computer interfaces that has become the ultimate goal of the BrainGate consortium: wireless, secure and fully automatic.
Portal "Eternal youth" http://vechnayamolodost.ru