Automated recognition of mental cognitive workload through electroencephalography and a deep learning approach

This study proposes a novel system for automatic cognitive workload identification from EEG signals. It overcomes limitations of traditional methods by integrating a pre-trained CNN with advanced features. The model processes a functional connectivity matrix, generated using the corrected imaginary phase locking value (ciPLV) method, with the Xception network. The CNN's output is then combined with nonlinear dynamic features. A feedforward neural network uses these combined vectors for classification, achieving high accuracy (over 98% for two workloads and 92.50% for three), demonstrating the promise of this integrated deep learning approach.