New Stanford invention lets paralyzed patients write with their minds

Scientists have found that when people want to do any action, make perception, speak, or do anything, the brain will produce brain waves with a specific pattern. For decades, scientists have been studying how to interpret these waves and connect them to machines so that the human brain can use the mind to direct the machine to do something.

Scientists call such a system a brain-computer interface (BCI). A study published May 12 in the journal Nature, which invented a system that can use the mind to type directly on a screen, is a cutting-edge achievement in the field of brain-computer interface research.

Some people are unable to type easily because of physical defects, while others are used to using voice input but are unable to speak for various reasons. This system can help them communicate with others easily.

The system, invented by a research group at Stanford University, implanted about 200 electrodes to read brain waves in the premotor cortex of a 65-year-old participant’s brain. Of course, not all brain waves are related to the information of the letters they want to read, so it is necessary to use a set of artificial intelligence software to interpret the brain waves, find the specific information, output the brain imaginary letters.

The researchers asked the participants to imagine the letters and words to write, the system automatically wrote them out on the screen, the equivalent of direct “mental typing”.

This research group had previously developed another similar system. That system allows the user to use the mind to move the cursor on the screen, through the cursor pressed on the screen on the keyboard to type letters. This system can only type an average of 40 letters per minute.

This new system uses artificial intelligence software to directly analyze the letters that the user wants to write, and the speed has increased a lot. Participants can type an average of 90 letters per minute, which is very close to the typing speed of the average person. The average average person types 115 letters per minute on a cell phone and 190 letters per minute on a keyboard.

The experiment demonstrated that the delay between the time the participant began to imagine the letters and the time they were typed on the screen was only about half a second. The error rate of the output message was only about 5%, and when used with auto-correction software, the error rate could be reduced to only 1%.

However, this is the result of experiments with pre-prepared text content. The researchers further tried to have the participants type with immediate intention, for example, answering some questions in this way.

In this case, the speed of typing was reduced to 75 letters per minute, and the error rate was still 2% after using with the auto-correct software. However, the researchers said such results are still acceptable, indicating that the system is still available.

The researchers say the system still has many imperfections. First, the experiment was conducted with only one participant, and it is not known if it works for everyone. Second, the set of letters interpreted by the system is relatively simple and does not yet include numbers, capital letters or various punctuation marks. Finally, these implanted electrodes reflect brain waves erratically and need to be calibrated about once a week.