Researchers have developed a new non-invasive brain-machine interface (BMI) that can be used to wirelessly control an electric wheelchair, robotic vehicle, or computing device by reading signals from the human brain.
While most brain-machine interfaces consist of unwieldy headgear that is loaded with electrodes and wires for scanning brain activity, this latest creation leverages the power of wireless sensors and compact electronics. The device is made of newfangled nanomembrane electrodes with flexible electronics and is paired with a deep learning algorithm that helps analyze the electroencephalography (EEG) signals.
Created by researchers at Georgia Institute of Technology, University of Kent and Wichita State University, the wireless BMI is comprised of pliable electrodes capable of making direct contact with the skin through hair and flexible circuitry with a Bluetooth telemetry unit. Electrodes are placed on the subject’s scalp, neck and below their ear, and they’re held in place with a fabric headband.
When EEG data is recorded from the brain, it’s sent to a tablet computer up to 15 meters away via Bluetooth. That’s when the deep learning algorithms come into play. The researchers noted that it’s challenging to minimize interference because the signals they’re working with are in the range of tens of micro-volts, which is similar to electrical noise in the body.
Deep learning is used to parse through that noise and drill down on the EEG signals that are most relevant for BMI purposes. This approach to filtering unwanted signals also contributes toward minimizing the number of electrodes required.
So far the system has been tested with six human subjects who have been able to control an electric wheelchair, a small robotic vehicle, as well as a computing device without using a keyboard or any other conventional controller. Going forward, the researchers aim to develop a method for mounting the electrodes on a hairy scalp without wearing a headband, as well as shrinking the electronics so more electrodes can be implemented in the same size package.
This research comes as countless startups and tech titans including Facebook are vying for a piece of the market. In Facebook’s case, the company announced in April 2017 that it was working on a method that would allow users to type with their minds at 100 words a minute, while more recently it spent an estimated $500 million to $1 billion on neural interface startup CTRL-Labs for its mind-reading wristband.