Playing Mind Games

When the topic of brain-computer interfaces (BCIs) comes up, most people think about brain implants and assistive technologies that can help those with severe motor disabilities to regain their independence. There is no question that this is a very important use of BCIs. By using these technologies, individuals can communicate naturally, control computers, and even drive wheelchairs with only their thoughts.

But the possible applications of BCIs extend far beyond assistive devices. Noninvasive sensing technologies exist that can measure brain signals without surgery, making them more appropriate for those without a disability. Using such devices, anyone can enjoy enhanced virtual reality experiences and more immersive entertainment options, or even control devices just by thinking.

In order for this technology to really catch on, for assistance, entertainment, or otherwise, it first needs to get quite a bit better. There are a number of areas that need to improve, like the sensing technologies and the algorithms that interpret those sensor measurements. But even as these areas continue to get better, another problem still plagues BCIs — calibration. Because every brain is different, the fact that a BCI works well for one user does not mean that it will work at all for another.

Accordingly, a BCI requires a calibration process to be carried out before a new user can make use of it. Unfortunately, this process is long, tedious, and costly. For assistive devices, this is a pain. But for other applications, it makes the technology completely impractical. Imagine having to spend a day at a research lab getting your new device calibrated before you can use it with your virtual reality headset!

A team at The University of Texas at Austin has been working to simplify this setup process to make BCIs more practical for everyday use. They developed a system that leverages a cap worn over the head that is packed with electrodes that noninvasively record electrical activity in the brain. The electrical signals are then forwarded to an artificial neural network that interprets the user’s intent, and uses that information to trigger any number of actions, like controlling a video game.

So far, this is not a novel idea. Many existing BCIs work in a similar fashion. But the team’s solution differs in that it is capable of self-calibrating to work with each user of the system. In addition to greatly simplifying the setup process, it also means that a number of individuals can share the device, as might be expected in a household.

The self-calibration process compares the brain signal patterns of the new user to those of previous users to quickly gain an understanding of how their brain signals differ. This comparison is done continuously and in real-time such that a new user can start using the BCI right away. The continuous adjustments made by this process help the system to get better over time, and can also help to account for changes in the user's brain activity as they age.

The team recruited eighteen healthy volunteers who were new to BCIs to participate in a five-day training program. The efficacy of the approach was tested in two different settings: a bar balancing task and a realistic car racing game. Throughout the training sessions, participants completed multiple runs of both tasks. The results showed that subjects were able to learn to use their BCI effectively after training in both tasks. Furthermore, it was observed that the improvement in BCI control was partly attributed to subjects acquiring the skill of producing more distinct brain signals, even if they did not match those of the original expert user that the algorithm was trained on.

At this time, the BCI has only been proven to work in binary tasks (e.g. turn left or turn right). Moving forward, the researchers will have to prove that multiclass problems can also be solved with a high level of accuracy. If so, this innovation could go a long way toward making BCIs more practical for everyday use.