A Database for Cognitive Workload Classification Using Electrocardiogram and Respiration Signal

Cognitive workload is a critical factor in determining the level of attentional effort exerted by users. Understanding and classifying cognitive workload is challenging as individuals exert varying levels of mental effort to meet the task's underlying demands. Twenty-six participants (12M, 14F, Mean = 22.68 ± 5.10) were exposed to two different tasks designed to induce low and high cognitive workloads. Subjective and objective measures were collected to create a novel, validated multimodal dataset for cognitive workload classification. Participants’ perceived workload was collected using the NASA-TLX. Electrocardiogram (ECG) and Respiration (RR) data were collected to extract the Heart Rate Variability and Respiration Rate Variability features. Four machine learning algorithms were utilized to classify cognitive workload levels where AdaBoost classifier achieved the highest Leave-One-Subject-Out Cross-Validation accuracy, and F1-Score of 80.2%, 80.3% respectively. This is the first publicly available dataset with ECG, RR and subjective responses for cognitive workload classification.