Researchers at Tufts University
in Medford, Massachusetts, want to give computers the ability to directly
monitor human brain in real time performance. The system utilises a headset
that beams infrared light from emitters on a user's forehead into their prefrontal
cortex, a part of the brain associated with planning and decision-making. Some
of the light is absorbed by oxygenated haemoglobin, some by the deoxygenated
version of the molecule, and some is reflected back out. By measuring the
amount of light reaching receivers on the forehead, the system can tell when a
user is concentrating intently or not mentally engaged. Matching the readings
to what a user is looking at on a screen allows the system to determine what
useful information is and what is getting in the way. The technique, known as
functional near-infrared spectroscopy (fNIRS), is a crude brain imager compared
with, fMRI. But infrared sensors are cheap and portable and MRI machines are
not. Researchers reckon they can glean enough information from their fNIRS rig
to turn computers into mind-readers.
As a proof of principle, the
system monitored haemoglobin changes while 14 test subjects rated movies listed
on the Internet Movie Database. It recorded how each user's brain behaved when
rating movies positively and negatively, with greater levels of activity
associated with more positive ratings. After this training, the system
recommended a list of other movies in turn, with each movie suggestion modified
by the brain's reaction to the previous movie suggestion. Not only were its
suggestions more acceptable than a random list, but it also improved its
results the more it was used. The US Federal Aviation Administration is also
exploring the technique to help manage the cognitive workloads of air-traffic
controllers. The next step is to build a brain interface that can handle more
complex interactions, like filtering emails and the other rivers of information
that threaten to overwhelm the modern worker on a daily basis. For now, their
set-up can only determine when people are engaged with what they are doing, and
when they are not.
More information: