Dr. Youngmoo Kim and Dr. Hasan Ayaz
Brain computer interfaces (BCIs) are systems that can potentially allow people with limited or no mobility to use a computer with only their thoughts. This proposed research is an initial step towards an assistive robot, capable of remote operation on behalf of the user. As a pilot study, 12 subjects have participated in an experiment to evaluate the use of motor imagery (imagined movements) of the hands and feet in a BCI. Brain activity is recorded using functional near infrared spectroscopy (fNIRS), a safe, portable, and relatively affordable optical brain-imaging system. With simple features and a linear discriminant analysis classifier, several subjects are able to attain 70% accuracy in a two-class left vs. right motor imagery problem. Additionally, the use of common average reference and task-related component analysis yield 40-50% accuracy for several subjects in a four-class problem.
This proposed work consists of three main contributions: evaluating the feasibility of a four-class motor-imagery-based fNIRS BCI; developing a real-time classifier that can be used to control the BCI; and proposing a follow-up study to implement control of a robot, as well as investigate the possibility of adding a resting state as a fifth class.