EMD-based Features for Cognitive Load and Stress Assessment from PPG Signals

Work-related stress and fatigue due to high-cognitive efforts have significant social impacts and are among the leading causes of health problems in modern work environments. We present a study on the applicability of Empirical Mode Decomposition (EMD)-based features for detecting the stress level and cognitive load from PPG signals. These features are extracted by decomposing the PPG signal with the EMD and the subsequent estimation of certain statistical parameters from the resulting intrinsic mode functions. Performance assessment of the proposed features is carried out in an experimental setup based on the publicly available CLAS dataset in a three-class decision problem considering different classifiers – Decision tree, kNN and SVM. An averaged classification accuracy of up to 87.8% is reported. We perceive the reported experimental results as promising and will intensify research to develop PPG-based stress detection applications.