Abstract
Acute mental stress activates the autonomic nervous system (ANS), modulating physiological parameters. To assess the ANS response, we collected multimodal physiological signals, including electrocardiogram (ECG), electrodermal activity (EDA), and respiratory activity from healthy participants. The experimental protocol was designed to induce a high stress level in one group (STRESS) and low stress in the other (CONTROL), undergoing the same cognitive tasks. Heart rate variability (HRV) indices, parameters from respiratory activity and EDA were computed and analyzed. First, the effect of the proposed stress manipulation on the ANS was assessed, showing that linear HRV and respiratory parameters significantly changed during cognitive tasks with respect to rest in both the groups, mainly when respiration activity was integrated in the analysis. Nonlinear HRV parameters and EDA-based indices presented more task-specific modulations. Significant differences among groups were found only for the mean RR interval and the EDA-derived parameters. Additionally, Random Forest models were trained, and feature importance was assessed through Shapley values. Results identified the amplitude of the phasic EDA component, respiratory sinus arrhythmia (RSA), HRV sample entropy, and mean breathing period as the features most clearly differentiating cognitive tasks from rest, highlighting the importance of a multimodal assessment of acute stress.