EEG index for control operators’ mental fatigue monitoring using interactions between brain regions

•An on-line index of operators’ mental fatigue is proposed.•It uses EEG signals from 6 regions of interest (ROIs).•It measures the changes in the EEG spatial covariance matrix.•It is highly correlated with an EOG index using Perclos.•It is in concordance with the Karolinska Sleepiness Scale scores.

Mental fatigue is a gradual and cumulative phenomenon induced by the time spent on a tedious but mentally demanding task, which is associated with a decrease in vigilance. It may be dangerous for operators controlling air traffic or monitoring plants. An index that estimates this state on-line from EEG signals recorded in 6 brain regions is proposed. It makes use of the Frobenius distance between the EEG spatial covariance matrices of each of the 6 regions calculated on 20 s epochs to a mean covariance matrix learned during an initial reference state. The index is automatically tuned from the learning set for each subject. Its performance is analyzed on data from a group of 15 subjects who performed for 90 min an experiment that modulates mental workload. It is shown that the index based on the alpha band is well correlated with an ocular index that measures external signs of mental fatigue and can accurately assess mental fatigue over long periods of time.