Cognitive Stress Detection during Physical Activity using Simultaneous, Mobile EEG and ECG signals

Stress is a psychological concept defined as an ensemble of coping responses to a perceived threat. It has received tremendous, well-placed scientific attention given its impact on individual health and performance. While acute exposure is associated with several physiological and psychological diseases; stress proved beneficial in some circumstances. It is therefore relevant to investigate it experimentally by inducing stress in participants through validated stressors that can elicit different forms of stress. To counter the personal bias embedded in questionnaires for stress assessment, physiological signals give access to a more objective, personalized response. Stress generates a wide range of reactions in terms of neurophysiological activity assessed using electroencephalography (EEG). Cardiac activity also proved relevant to analyze changes in heart rate and rhythm in electrocardiography (ECG) signals. In this paper, we propose a data-driven approach for differentiating stress from physiological baseline based on the multi-modal PASS database. Data is from mobile, simultaneous recording of EEG and ECG data, in different settings combination of stress elicitation and physical activity intensity. The method leverages the use of Limited Penetrable Visibility Graphs (LPVG) by extracting various features from images of adjacency matrices of the signals; including frequency-based and shape-based features. These features were then input into different machine-learning models. The proposed approach's performance was rigorously evaluated using real data. The obtained results provide compelling evidence supporting the feasibility and effectiveness of the proposed method for stress detection.