EEG-based mental workload analysis for multitasking testing and training systems

Abstract

The mental workload and multitasking capacity of an individual is an important consideration for operator and workplace safety assessment. Developing ways to understand and accurately assess mental workload is therefore essential. In this thesis, we study multitasking mental workload elicited from the simultaneous capacity (SIMKAP) psychology test, measured with the electroencephalograph (EEG) signal. Beginning with the individual specific case, we propose novel feature based methods for classification, drawing inspiration from previous work and psychophysiology. We then study general underlying neural mechanics of multitasking through EEG spectral analysis of the SIMKAP test and propose a subject independent classification model based on questionnaire ratings. Next, we consider generalization capability by transferring models trained on SIMKAP to classify a separate workload dataset and show that a novel 2-level autoencoder structure is able to learn features for stable transfer classification performance. Finally, we show applications developed for multitasking assessment and neurofeedback training.