Abstract

Mental workload can be described as individuals' cognitive exertion during task performance, where an employee may end up being mentally overloaded by the time it is evening. In this paper, we present a technique for forecasting evening mental workload index from the EEG signals recorded in the morning, given a linear relationship between morning and mental workload. The multilayer perceptron neural network is designed to forecast the evening mental workload index using higher-level signal features extracted from alpha and theta EEG signals using Convolutional Neural Network. Methodology for the EEG signal transformation to images for feature extraction are key novelties presented in this paper. We thus effectively established a forecasting relationship capable of determining and predicting an individual's mental workload. During our research we used EEG data recorded from university lecturers and university staff. Findings demonstrate the suitability of our model, as there is no evidence suggesting the existence of a non-linear relationship- we are thus able to use the proposed approaches and model for practical implementation in day-to-day occupations.