Unsupervised Short-term Covariate Shift Adaptation for Self-paced BCI

one of the major challenges in Brain Computer Interface systems (BCIs) is dealing with non-stationarity in EEG signal. There are two types of nonstationarity in EEG signal: 1)long-term changes related to fatigue, changes in recording conditions or effects of feedback training which is addressed inclassification step and 2) short-term changes related to different mental activities and drifts in slow cortical potentials which can be addressed in the feature extraction step. In this paper we use acovariate shift minimization method to alleviate short-term (single trial) nonstationarity effects of EEG signal and improvethe performance of the self-paced BCIs in detecting foot movement from the continuous EEG signal. The results ofapplying this unsupervised covariate shift minimization with 2 different classifiers, linear discriminant analysis (LDA) and probabilistic classification vector machines (PCVMs) and with two different filtering methods show the considerable improvement in system performance

Keywords:

Self-paced Brian Computer interfaces , Nonstationarity , feature shift minimization