EEG-based mental workload estimation using deep BLSTM-LSTM network and evolutionary algorithm

The mental workload can be estimated by monitoring different mental states from neural activity. The spectral power of EEG and Event-Related Potentials (ERPs) are the two mediums for monitoring the mental states. In this paper, we estimate the workload during the multitasking mental activities of human subjects. The estimation of mental workload is done using the "STEW" dataset [16]. The dataset consists of two tasks, namely "No task" and "simultaneous capacity (SIMKAP)-based multitasking activity". Different workload levels of two tasks have been estimated using the composite framework consists of Grey Wolf Optimizer (GWO) and deep neural network. GWO has been used to select optimized features related to mental activities. Other optimization techniques such as Genetic Algorithm (GA), Particle Swarm Optimization (PSO) are generally slower compared to the convergence rate of GWO. A deep hybrid model based on Bidirectional Long Short-Term Memory (BLSTM) and Long Short-Term Memory (LSTM) has been proposed for the classification of workload levels. The proposed deep model achieves 86.33% and 82.57% classification accuracy for "No task" and "SIMKAP-based multitasking activity," respectively. A judicious distinction between different workload levels at higher accuracy will essentially increase the performance of an operator, which effectively improves the efficiency of the Brain-Computer Interface (BCI) systems. (C) 2020 Elsevier Ltd. All rights reserved.