Humans have long been fascinated by the notions of people and something that can read our minds, such as telepathy, computers, and Santa Claus. Researchers now say they can develop a system that combines machine learning with the brain-computer interface (BCI) to read writing that occurs in the brain, not on paper.
A man paralyzed through a computer interface can “input” 90 characters per minute. According to a new study, those who are using the ‘brain-to-text’ system have the potential to leap communication. Rather than trying to enable a virtual keyboard by reading neural activity in the brain, the groundbreaking team of responsibility focused on virtual handwriting tracking. Conventional systems that track brain activity and map it to computers generally think about arm movements. By tracking them, you can map them to a virtual keyboard or other kinds of interface highlighting keys, even if the arm itself cannot be moved.
While it works, the speed is limited. According to Krishna Shenoy, a researcher at Stanford University’s Howard Hughes Institute of Medicine and who wrote this new study with Stanford’s neurosurgeon Jamie Henderson, in the current system, about 40 characters are possible every minute via the Brain-Computer Interface (BCI). When people imagine their hands, they see brain activity instead of arm movement. This is a similar brain implant technology that Elon Musk-backed startup Neuralink is working on, to allow paralyzed people to type and communicate efficiently without using their hands.
According to Frank Willett, a neuroscientist who participated in the HHMI research project, imagining handwritten characters presents a very unique pattern of activity. Algorithms trained to recognize this are much faster than traditional BCI systems.
A 65-year-old man had a grid of 2 small electrodes buried on the surface of the brain. Electrodes register the electrical activity of the parts of the brain which control the movement of hands and fingers. The man was paralyzed under the neck but imagined writing a letter with his hand. Using the following algorithm, the researchers understood the neural patterns associated with each conceived character and converted those patterns into text on the screen.
Participants generated 90 letters or 15 words per minute as well as brain activity. Researchers and colleagues at Stanford University’s Howard Hughes Institute of Medicine reported to Nature on May 12th. This is almost the same as the average typing speed on a smartphone before and after the participant’s age.
Thought the text system was working even after getting hurt. Shenoy said that the big surprise is that after a spinal cord injury or a few years after the inability to use the hands and fingers, we can still hear its electrical activity. It’s still very active. Thought-based communication is still in its infancy. We need more volunteer research, but “there is no doubt that it will work for others again,” says Shenoy. Researchers will test the system on people who have lost all their ability to move and speak.