Evaluation of Working Memory Load using EEG Signals

This paper investigates mental workload assessment using statistical features derived from electroencephalography (EEG) signals. Mean, root mean squared, and correlation-based features are extracted from data including EEG signal recordings of five participants performing a reading task with three difficulty levels of low, medium, and high and a baseline condition. Results reveal that for the given task, features derived from the EEG signals consistently exhibit a very high degree of discrimination between the induced load levels, confirming EEG as an important method for the real time, objective determination of cognitive load level. Also, the frontal EEG channels appear to be sensitive to the working memory load than the other channels. Analysis of the effect of the window length used during feature extraction from the EEG signals suggest that features extracted from EEG segments as short as 1 second exhibit an acceptable amount of standard deviation, suggesting that EEG-based measurement of fairly rapid changes in cognitive load may be feasible.