Electrodermal activity (EDA) and heart rate variability (HRV) offer opportunities to grasp critical manifestations of the nervous autonomic system using low-intrusive sensing tools. A key question relies on the capacity to adequately process EDA and HRV signals to extract cognitive load markers, a multifaceted construct with intricate neural networks functioning, where emotions interfere with cognition. Here, 34 participants (20 males, 19.2 ± 1.3 years) were exposed to two-back mental tasking and watching emotionally charged images while recording EDA and HRV. HRV signals were processed using variable frequency complex demodulation (VFCDM) and wavelet packet transform (WPT) to provide high- and low-frequency (HF and LF) markers. Three methods were used to extract EDA indices: VFCDM (EDATVSYMP), WPT (EDAWPT), and convex-optimization (EDACVX). Cognitive load and emotion epochs were distinguished by significant differences in NASA-TLX scores, mental fatigue, and stress, on the one hand; and by EDACVX and, remarkably, EDATVSYMP and HF-HRVVFCDM on the other hand. A linear mixed-effects model and stepwise backward selection procedure showed that these two markers were main predictors of the NASA-TLX score (cognitive load). The individual perception of cognitive load was finally discriminated by k-means clustering, showing three profiles of autonomic responses relying, respectively, on EDATVSYMP, HF-HRVVFCDM, or a mix of these two markers. The existence of EDA-, HRV-, and EDA/HRV-derived profiles might explain why previous attempts that have predominantly employed a single biosignal often remained unconclusive in evaluating the perceived cognitive load, thereby demonstrating the added value of the present approach to monitor mental-related workload in human operators.Read less <