How can a computer understand someone’s intention to move something and convert it into a technical signal?
Müller: For communications between human subjects and the computer, we use a traditional EEG unit. With 128 electrodes attached to the skin of a subject’s head, we record brain waves. The intention to move the right or left hand corresponds to a specific pattern in our brain activity. The interface filters the pattern out of the variety of data with an intelligent signal-processing and classification algorithm and converts it into a control signal.
What can a test subject attached to a BCI do with the power of thought?
Müller: To motivate test subjects during long lasting tests, we have connected a simple computer game, Brain Pong, to the BCI. Test subjects can simply imagine moving their right or left hands to control a bar and play “teletennis.” They can also steer a cursor onto specific letters and thus speak without opening their mouths. So far, we’ve been able to convert only very simple impulses, such as “right or left,” “above or below,” or maybe even “on or off.” Commands such as “open door” are just too complex right now.
Right now, how long does the BCI take to convert the thought, “I want to move my hand,” into a visible movement?
Müller: Forty milliseconds.
That’s rather fast. What impulses require more time?
Müller: The more complex the sample movement, the longer the initial training time. Once the sample has been classified, the conversion of any impulse into visible movement conditioned by the hardware lasts just as long: about 40 milliseconds.
What project goals do you have for the upcoming years?
Müller: First of all, we want to make the device faster so that it’s truly suitable as a means of communication. At the beginning of our research about three years ago, we could transmit only five bits per minute. Now we’re up to 37, and that’s world-class when compared internationally. Nevertheless, traditional mice transmit about 300 bits per minute. Second, we want to simplify the hardware. Right now it takes about an hour to place the EEG electrodes properly. Our dream is a non-contact EEG that’s integrated into a sort of baseball hat and that we can simply put on a subject’s head.
An additional vision for the future is that those who suffer from serious illnesses can use the BCI to communicate with their environment once again and to control prostheses without having to go through the effort of training other muscles. Maybe our cars will also be safer because they can recognize an abrupt movement, such as quickly turning the steering wheel or slamming on the brakes early, and react appropriately. And of course, computer gamers could replace their joysticks with a cool cap – if they want to.
What characterizes the Fraunhof BCI program in comparison with previous research on human-machine interaction?
Kaplow: Previous research on this topic has assumed that humans must adjust themselves to the machine. Our motto is, “Let the machines learn.” That means that our algorithms are adjusted to the brain waves of the subjects and that they continuously learn to filter and classify them better. The actual learning is done by the computer. This approach has let us reduce the previously typical training time for subjects from several hundred hours to 20 minutes.
Can the subject who transmits an impulse influence how the computer analyzes the brain signal?
Can a single brain signal lead to different movements, depending on the computer program that analyzes it?
Müller: No. Each movement has exactly one representation in the brain, and the representation is converted into zeros and ones in the computer. The computer has no option for interpretation – no choice between alternatives.
Now people are going to worry that the program will evaluate a brain impulse differently than that subject intended. Is anything like this conceivable?
Müller: No, because the computer doesn’t receive anything other than the impulse that the subject intended to transmit.
Should we be afraid of being manipulated by this technology?
Kaplow: No. Communication between a human and the machine is a one-way street.