EMD-based discrimination of mental arithmetic tasks from EEG signals

Recognition of mental arithmetic activities from electroencephalogram (EEG) signals forms an important constituent in developing brain-computer interface (BCI) systems. In this paper, we explore an approach based on empirical mode decomposition (EMD) to identify different arithmetic activities from EEG signals. The EMD is a technique proficient at processing non-stationary signals, such as EEG signals, by decomposing into different frequency components referred as intrinsic mode functions (IMFs). Eight attributes are computed from each IMF to obtain the feature vector for a signal. These features are then passed to support vector machine (SVM) classifier for the recognition of normal EEG signal and the signal recorded for mental arithmetic tasks. Using 10-fold cross-validation technique, the proposed method accomplished an average classification accuracy of 95%, which is better than the current approaches explored in the literature.