Photoplethysmography-Based Assessment of Mental Stress: Stroop Test Analysis Using Public and IoT Data

Photoplethysmography (PPG) serves as a non-invasive technique for assessing an individual’s physiological state, including stress. Cognitive tasks typically alter this state compared to a pre-test baseline. Specifically, the Stroop test is recognized for modifying sympathetic nervous system activity and ventricular repolarization dynamics, making PPG monitoring a suitable tool for tracking these variations. However, there is a lack of efficient, monomodal methods for stress detection, as existing approaches still depend heavily on traditional heart-rate-variability features. For the experimental evaluation, a dataset of 170 PPG signals was aggregated from different sources. Cognitive load detection is performed using a BootStrap Aggregation approach and features are ranked and selected by Random Forest and Robust Rank Aggregation methods. Moreover, the hyperparameters of the classification model are tested across different feature dimensions. By evaluating a heterogeneous feature set, six top-ranked entropy measures forming the feature vector are selected. The proposed PPG-based detection exhibits a high-level performance for both publicly available and mixed-source data. Monomodal detection could offer efficient and practical insights for addressing stress-related physiological changes.