Severely disabled people who cannot operate a motorised wheelchair may one day get their independence, thanks to a system that lets them steer a wheelchair using only their thoughts.
Unlike previous thought-communication devices, the system does not use surgical implants. Instead a skullcap peppered with electrodes monitors the electrical activity of its wearer鈥檚 brain. Early trials using a steerable robot indicate that with just two days training it is as easy to control the robot with the human mind as it is manually.
鈥淚t鈥檚 a very positive step,鈥� says Paul Smith, executive director of The Spinal Injuries Association in London. 鈥淭he psychological benefits it would offer are huge.鈥�
Advertisement
The current options to give freedom of movement to people who are quadriplegic are limited, says Smith. For example, it is possible to steer a wheelchair using a chin-operated joystick or by blowing into a thin tube. But both options can be exhausting 鈥� and they are not suitable for those with very limited movement.
So Jos茅 Mill谩n at the Dalle Molle Institute for Perceptual Artificial Intelligence in Martigny, Switzerland, along with researchers from the Swiss Federal Institute of Technology in Lausanne and the Centre for Biomedical Engineering Research in Barcelona, Spain, has come up with a system that can reliably recognise different mental states.
If all goes to plan, it will be the first mind-controlled system able to operate something as complicated as a wheelchair, says Mill谩n.
At the moment the system controls a simple wheeled robot. The user dons the electrode-lined skullcap, which monitors electrical activity on the surface of the head. A web of wires sends the information to a computer. Mill谩n鈥檚 software then analyses the brain鈥檚 activity and, using a wireless link, passes on any commands it spots to the robot.
At the moment the user can choose between three different commands: for example, 鈥渢urn left鈥�, 鈥渢urn right鈥� and 鈥渕ove forward鈥�. Mill谩n鈥檚 software exploits the fact that the desire to move in a particular direction will generate a unique pattern of brain activity. It can tell which command the user is thinking of by spotting the telltale pattern of brain activity associated with that command.
To ensure the robot does not hit any objects, it contains some inbuilt intelligence. So, when the user thinks of one of the three states 鈥� for example, 鈥渢urn left鈥� 鈥� the software translates it into an appropriate command for the robot, such as 鈥渢urn left at the next opportunity鈥�.
In this case, infrared sensors allow the robot to detect walls and objects and it will safely plod along until it reaches the next turning. And in case the software has got the command wrong, a light on the robot indicates what it is going to do, giving the user time to correct it.
Mill谩n鈥檚 skullcap-centred system is a significant step forward. Five years ago, surgeons in Atlanta, Georgia, grabbed the headlines by implanting electrodes in the brain that allowed patients to communicate by controlling a cursor on a computer screen (New 杏吧原创 print edition, 17 October 1998, p 5).
But the risks associated with such invasive methods mean approaches such as Mill谩n鈥檚 that use electroencephalography (EEG), in which surface scalp electrodes monitor electrical activity, are preferable.
However, methods using current EEG technology are slow to recognise different mental states and can only do so by measuring the brain鈥檚 alpha waves. Since this involves shutting the eyes and relaxing, it is not a practical option for people trying to control a wheelchair.
So the team has designed its own software to analyse the activity from a standard eight-electrode EEG array. It uses a neural network that can be trained to recognise complex non-alpha-wave patterns and relationships more quickly. This means Mill谩n鈥檚 system works in real time. 鈥淵ou can identify any pattern you think needs to be translated into a physical action immediately,鈥� he says.
The team is now trying to increase the number of mental states that its system can recognise. 鈥淭he larger the number of mental states you have the more complicated it becomes,鈥� Mill谩n says. 鈥淲e now need to improve the learning algorithm in order to differentiate the EEG patterns.鈥�
Another grey area, according to Mill谩n, is whether there will be a drop in the quality of the EEG signals when the user is actually sitting in the chair as it moves. It is possible that a person鈥檚 brain activity will become a lot noisier as they take in their moving surroundings, he says.