Linear decoding of 2D hand movements for target selection tasks using a non-invasive BCI system

In this paper, a visual interface showing target selection tasks has been designed to decode 2D velocities from a brain-computer interface (BCI) system during hand movements. Ten healthy volunteers were asked to perform hand movements using a mouse to control a cursor on the screen. Three targets were randomly highlighted and the volunteers were asked to reach them with the cursor. The actual position of the cursor and the EEG activity were simultaneously acquired. To decode hand velocity, the EEG signals registered from 16 electrodes were processed and linear models were used to obtain the transformation to velocity values of the hand movement. The results of the decoded velocity were compared to the actual values using the Pearson correlation coefficient showing average correlation coefficients of 0.39 in the X-axis and 0.47 in the Y-axis, showing an improvement over previous results obtained when applying similar techniques. It is also relevant that the number of electrodes used has been significantly reduced (from 34 to 16) and the use of a 2D environment shows a great potential in future rehabilitation systems for stroke patients.