Researchers from Imperial College London have developed a brain-computer interface that gives a person the ability to control additional robotic arms without interrupting the natural movements of their own hands. The results were published in the journal Nature Communications.
The system is based on the brain's response to vibration, the so-called P300 response. This is a burst of electrical activity that occurs roughly 200-500 milliseconds after a person notices a rare or important signal among many others. In the experiment, participants wore four small vibrators on their forearms. Each vibrator corresponded to one of four commands and was switched on in random order for 200 milliseconds with pauses. To select a command, the person mentally focused on the vibration at the desired point, ignoring the others. A special machine-learning program (a computer algorithm that finds patterns in data) detected the moment the P300 wave appeared and matched it to which vibrator had just fired.
The study involved ten healthy right-handed volunteers aged 23 to 37 (six men and four women), eight of whom had no prior experience with neural interfaces. Each wore a cap with 64 sensors to record electrical brain activity (electroencephalography). The scientists tested whether such a system could recognize commands when both hands were occupied with a complex task - holding a tray with a ball.
On the first day, command recognition accuracy while simultaneously performing the tray task was 53.1% versus 68.1% at rest. On the second day, the gap persisted (56.3% versus 77.2%). However, by the third day, accuracy under multitasking conditions rose to 69.7%, and the difference from the resting result (79.1%) was no longer statistically significant. In other words, after several training sessions, the participants learned to handle the interface almost as well as without the distracting load.
The authors attribute the improvement to the brain responding more consistently to vibration over time: the spread in timing between individual P300 bursts decreases. This allows the program to detect the signal against background noise more accurately. At the same time, the hand movement itself reduced the amplitude of the brain response - on average to 4.7 microvolts versus 7.2 microvolts at rest on the third day. The scientists suggest that practice helps people distribute attention better, partially compensating for the natural dampening of sensory signals during movement.
Importantly, working with the interface did not impair hand coordination: the speed and accuracy of rotating the ball on the tray did not decrease. In an additional demonstration test, one participant used the interface to control two desktop robotic arms that performed predetermined movements - for example, one arm lifted a cup while the other threw a straw into it.
However, the approach has limitations. To reliably recognize a command, the system requires several vibration cycles, so the response takes seconds rather than fractions of a second as with a living hand. The robotic arms performed programmed actions rather than being smoothly controlled by the user. In addition, prolonged vibration can cause skin habituation and reduce the brain's response. Further trials with more participants and faster algorithms will show how applicable the technology is outside the laboratory.