Robotic systems can improve daily life for over one billion people with disabilities worldwide. Brain-computer interfaces, or BCIs, are devices that let the brain communicate directly with machines, skipping the need for muscle movement. This is helpful for people who cannot move easily. While some BCIs require surgery to place devices inside the brain, these invasive methods are risky and only used for serious medical cases.
Bin He, a professor at Carnegie Mellon University, has worked for over 20 years on noninvasive BCIs, which do not need surgery. His research uses electroencephalography, or EEG, a method that records brain activity through sensors on the scalp. His group has made important achievements, such as flying a drone, moving a robotic arm, and controlling a robotic hand using only brain signals.
Breakthrough in robotic hand control
A recent study published in Nature Communications shows progress in using EEG-based BCIs to control robotic hands. The study focused on decoding brain signals to move individual fingers on a robotic hand in real time. This is important because hand movement is essential for daily tasks, and even small improvements can greatly enhance quality of life. However, EEG signals are not very precise, making it hard to control specific finger movements.
In this study, researchers used a new deep-learning method to interpret brain signals. They also used a fine-tuning technique to improve how the system reads EEG signals continuously. Human subjects thought about moving their fingers, and the robotic hand copied those movements, performing tasks with two or three fingers. This was the first time EEG-based BCIs achieved such precise control without surgery.
The next goal is to improve this technology for more complex tasks, like typing. This work could make noninvasive BCIs more useful for many people, not just those with severe disabilities. The study shows that EEG-based BCIs have great potential for detailed movement control, which could transform how people interact with robotic devices.