All in the mind

researchers are developing a system that could give people completefreedom to move a cursor around a screen using thought alone.

The research takes a novel approach to brain computer interfaces (BCI) that gives a person's electroencephalograms (EEGs) complete control of a computer.

By the end of the project the team, headed by Dr Francisco Sepulveda, hopes to have created a system that dispenses completely with the mouse/keyboard interface. Joysticks, haptic interfaces, voice recognition and myolectriccontrols could be rendered obsolete, too.

'The challenge will be to make the brain-driven interface as reliable as these', said Sepulveda. 'This project will investigate how close we can get to this goal by means of genetic algorithms and better visual interface design.'

Users of the system include, but are not restricted to, people with disabilities or severe musculoskeletal disorders.

Current BCI designs work by observing EEG signals, then looking for specific features or waves, manipulating them and then making a decision as to whether such signals are meant to move a cursor up and down or left and right. This binary approach is achieved using a keyboard-like or 'daisy wheel' layout. The user thinks of a character and this is classified as 'on', switching all other characters to an 'off' position.

'Previously people have used a binary mode of control where you go in one direction or the other and that's it, so the margin of error was very large,' said Sepulveda.

He added that the team would not use this approach. Instead, it will determine to what degree the user is interested in all directions (among at least four choices) and a linear combination of these is used. This analogue approach allows the BCI-mouse to move in the same way as as current mice, and not just in steps based on a small set of possibilities.

Instead of requiring extensive training, users will go through a two-hour calibration process.

Human EEG signals are unique and change on a regular basis, even when the same thought process occurs. A person thinking 'move the cursor left' 10 times may give off 10 similar but slightly different EEG signals. The new system will use evolutionary algorithms that adapt to each individual's EEG patterns. These algorithms discover the most successful filters to use to find the best signal to prompt a command.

By placing the training burden on the computer rather than the user, the team hopes to make the system more appealing.

While the technology could have a huge impact on the lives of disabled people, Sepulveda believes that it will only take off once the private sector gets behind it. With this in mind the team is aiming to create a system that is easy to use and dispenses with the usual invasive techniques employed to pick up EEG signals.

Currently, EEG data is collected through electrodes placed in a cap and conductive gel connects them with the user's skin. To speed up the system, much of the processing is being done at the electrode level using head stage amplifiers and filters so the data passed to the computer has already been filtered.

While the researchers hope that in the future dry electrodes will be able to be used without the gel, they believe their current approach is preferable to using electrodes that require needles to connect them with the scalp.

'We have decided to stick to two things to take this technology further,' said Sepulveda. 'One is to not require extensive training on the user side and the other is to make sure we always use non-invasive technology. We know that by doing those things we will not explore the technology to its fullest potential, but we feel we can take it a long way and if we want to appeal to the public we must take these things into consideration.'

Although the team has made great progress it believes a final product could still be up to 10 years away. It is now trying to increase the reliability of the system, as well as the speed, which so far has lagged behind current BCI systems.

'We want to try and see how we can improve on the visual interface so that we get better signals, increase the reliability and lessen the burden on the user,' said Sepulveda.

'Also one main goal is to try to bring this technology out of the lab and do business with companies that might want to try it at the consumer level. I do believe the biggest advance in this area will come when virtual reality and gaming industry partners join us in our research.'

Sepulveda and the BCI group recently developed a prototype mouse capable of full 2D motion control that can be used within minutes and without training by anyone that has been through the calibration process. This June the team will use a £370,000

grant to take their research further forward.