Binary EEG Control for Two-Dimensional Cursor Movement: An Online Approach

Electroencephalography (EEG) is an appealing basis for brain-computer interface technology because EEG is non-invasive. However, because EEG signals are spatially blurred and typically have very low signal-to-noise ratios, extracting relevant information in the single-event case is challenging. The most easily accessible information is one-dimensional (for example, mu rhythm amplitude, average hemispherical power, or presence of a P300 evoked potential). Many studies have attempted to use such one-dimensional parameters as a basis for control. Robust results may be obtained when control is restricted to answering "yes" or "no" questions, such as comparison of a value to a threshold. However, possible applications of such control have been limited, and more dimensions of control are desirable. This research presents a new technique for obtaining more dimensions of control from existing technology. Yes/no answers are taken sequentially in groups of n, and in combination designate a specific choice from 2^n possible values. This is homologous to the function of bits, and consequently has been termed "binary control." To demonstrate this approach, a two-dimensional cursor control paradigm was developed in MATLAB. Users move a cursor among squares of a grid towards a target while avoiding a trap. At each move, there are up to four positions into which the cursor may be directed (up, down, left, and right). In this embodiment, control is achieved by twice comparing average alpha-and beta-frequency power of each hemisphere during continuous imagined lateralized hand movement. The first comparison narrows the four choices to two, and the second uniquely determines the cursor movement. This paradigm was shown to be compatible with the Brain-Computer Interface-to-Virtual Reality (BCI2VR) software, and preliminary tests were run on normal volunteers. These tests demonstrated the feasibility of pursuing future research with binary control. Binary control is promising because of- its robust underlying principles, and because it is easily expandable and adaptable. The source of control may be any EEG feature that can signal a yes/no answer, and the quantity of possible choices doubles with the addition of each answer "bit." This might provide means for more complex control, such as of a robotic arm or virtual keyboard. The binary approach might also prove more efficient than current EEG-based control methods, possibly with less computational demand.