Visual Mental Workload Assessment from EEG in Manual Assembly Task

Abstract

The use of electroencephalography (EEG) to assess mental workload (MWL) has been the subject of many studies. Also, there have been many efforts to achieve task-independent MWL estimation, with the most recent being in the field of machine learning (ML). However, the estimation still remains highly dependent on the specific task used for ML model training. Furthermore, there is a shortage of research that is focused on developing an estimator that would function for multiple different tasks within a specific task domain. The creation of the dataset described in this work is a step towards developing task-independent ML estimator within the scope of visual cognition. An experiment meant for the ML model training is designed to collect EEG signals for different levels of MWL during manual assembly that involves assembly instructions to be visually processed by operators. It includes idle state of an operator, as well as two different complexity levels of the visual instructions. EEG data is collected using wireless EEG-recording cap that can be easily incorporated in everyday assembly line environments.