EEG-based workload estimation technology provides a real time means ofassessing mental workload. Such technology can effectively enhance theperformance of the human-machine interaction and the learning process. Whendesigning workload estimation algorithms, a crucial signal processing componentis the feature extraction step. Despite several studies on this field, thespatial properties of the EEG signals were mostly neglected. Since EEGinherently has a poor spacial resolution, features extracted individually fromeach EEG channel may not be sufficiently efficient. This problem becomes morepronounced when we use low-cost but convenient EEG sensors with limitedstability which is the case in practical scenarios. To address this issue, inthis paper, we introduce a filter bank common spatial patterns algorithmcombined with a feature selection method to extract spatio-spectral featuresdiscriminating different mental workload levels. To evaluate the proposedalgorithm, we carry out a comparative analysis between two representative typesof working memory tasks using data recorded from an Emotiv EPOC headset whichis a mobile low-cost EEG recording device. The experimental results showed thatthe proposed spatial filtering algorithm outperformed the state-of-thealgorithms in terms of the classification accuracy.
展开▼