Localization of Brain Activity in Electroencephalography Data during Brain-Computer Interface Operation

Abstract: In this Master's thesis I present a means of finding active sources of cortical electrical activity from electroencephalogram (EEG) data acquired during operation of a brain-computer interface (BCI). A novel subspace-based technique was used to suppress spatially correlated EEG interference sources, followed by a technique that estimates the source parameters with a near maximum likelihood performance. These sources are found to correlate with event-related potentials (ERPs) and are thus hypothesized to be responsible for the N200 and P300 ERPs. The source localization technique was tested on EEG data of 6 able-bodied subjects, and my analysis underlines consistencies and variation of brain activity locations both within and across subjects. Results are compared to literature and results using other techniques and the new methods show promise in localizing brain activity when dual-condition datasets are available.