Physiological-signal-based mental workload estimation via transfer dynamical autoencoders in a deep learning framework

Abstract

Evaluating operator mental workload (MW) in human-machine systems via neurophysiological signals is crucial for preventing unpredicted operator performance degradation. However, the feature of physiological signals is associated with the historical values at the previous time steps and its statistical properties vary across individuals and types of mental tasks. In this study, we propose a new transfer dynamical autoencoder (TDAE) to capture the dynamical properties of electroencephalograph (EEG) features and the individual differences. The TDAE consists of three consecutively-connected modules, which are termed as feature filter, abstraction filter, and transferred MW classifier. The feature and abstraction filters introduce dynamical deep network to abstract the EEG features across adjacent time steps to salient MW indicators. Transferred MW classifier exploits large volume EEG data from an source-domain EEG database recorded under emotional stimuli to improve the model training stability. We tested our algorithms on two target EEG databases. The classification performance shows TDAE significantly outperforms existing shallow and deep MW classification models. We also investigated how to select TDAE hyper-parameters and found its superiority in accuracy can be achieved with proper filter orders.