SMOLT: Sensitive Mental Workload Assessment Enhanced with Multi-Task Learning

Mental workload is considered one of the most important contributors to human performance. In previous decades, considerable research has been conducted on workload assessment using different methods, such as subjective measurement and performance measurement. Recently, there is a trend to utilize physiological parameters, such as Electroencephalography (EEG) and Electrocardiography (ECG), for automatic objective workload assessment. Although significant progress has been made on physiological parameter based workload assessment, there are still a number of challenges to be addressed for automated workload assessment, two of which are: workload assessment in multiple dimensions (such as visual and cognitive) and high performance assessment. In this research, we propose a SMOLT software tool (sensitive mental workload enhanced with multi-task learning) for multidimensional workload assessment. SMOLT is innovative to build a multi-dimensional workload assessment model by incorporating advanced Multi-Task Learning (MTL) theory and multimodal deep learning, which is unique to model the relatedness among the output tasks (workload in different dimensions) and among input signals (multimodal deep learning for better feature representations). The proposed SMOLT is built on a significant amount of researches by our team to cognitive state assessment. We will incorporate the existing algorithms into SMOLT software and enhance the workload assessment in multi-dimensions.