Most people suffering from multiple sclerosis or spinal cord injuries can still move their eyes because they are directly connected to the brain. Some existing technologies already allow severely disabled people to stare at arrows on a computer and direct the movement of a wheelchair.
But there are problems with that system, including a delay between the movement of the eyes and the wheelchair.
"Current tracking software often uses a screen-based system where you have a screen open and you look at locations on the screen. The problem with that is that it's very simplistic and also diverts the users' attention from the outside world and therefore there's more risk of not noticing obstacles or other things in the way," said Kirubin Pillay, a PhD student at Imperial College London.
A team led by Aldo Faisal at Imperial College London has developed software that allows users to maneuver the chair just by looking in the direction they want to take.
"Our eyes are not only a window into our soul, they're also a window to our intentions. So if you want to go somewhere, for example if I want to go there, or go there, I will look there and I will look there in a specific manner, and we can build a computer system that can decode our eye movements, and so we observe eye movements with an eye tracker, and we then try to make sense of them, and the computer interprets these commands and drives the wheelchair accordingly,” said Fasial.
Two cameras trained on the eyes observe their movements and can determine whether a patient is merely looking around or wants to move in a certain direction.
"So essentially we track the pupil of the eye and via a calibration process, we relate that to where the subject's looking in the world around them," explained William Abbott, a researcher at Imperial College London.
Visual information detected by cameras is analyzed by algorithms within 10 milliseconds and translated into instructions for movement that's almost instantaneous.
The camera-based system costs only about $85 because most of the work is done by the algorithms. No expensive hardware is needed. The London team hopes to make the system commercially available within three years.
