Boffet Alexis, Arsac Laurent M, Ibanez Vincent, Sauvet Fabien, Deschodt-Arsac Véronique
Laboratoire IMS, CNRS, UMR 5218, Université de Bordeaux, Talence, France.
Thales AVS FRANCE SAS, Mérignac, France.
Sensors (Basel). 2025 Apr 8;25(8):2343. doi: 10.3390/s25082343.
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 (EDA), WPT (EDA), and convex-optimization (EDA). Cognitive load and emotion epochs were distinguished by significant differences in NASA-TLX scores, mental fatigue, and stress, on the one hand; and by EDA and, remarkably, EDA and HF-HRV 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 EDA, HF-HRV, 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.
皮肤电活动（EDA）和心率变异性（HRV）为使用低侵入性传感工具来掌握自主神经系统的关键表现提供了机会。一个关键问题在于能否充分处理EDA和HRV信号以提取认知负荷标记，认知负荷是一个多方面的概念，其神经网络功能复杂，且情感会干扰认知。在此，34名参与者（20名男性，年龄19.2±1.3岁）在记录EDA和HRV的同时，接受了双任务心理任务并观看了充满情感的图像。HRV信号通过可变频率复解调（VFCDM）和小波包变换（WPT）进行处理，以提供高频和低频（HF和LF）标记。使用三种方法提取EDA指标：VFCDM（EDA）、WPT（EDA）和凸优化（EDA）。认知负荷和情感阶段一方面通过NASA - TLX分数、精神疲劳和压力的显著差异来区分；另一方面通过EDA，尤其是EDA和HF - HRV来区分。线性混合效应模型和逐步向后选择程序表明，这两个标记是NASA - TLX分数（认知负荷）的主要预测指标。最终通过k均值聚类区分了个体对认知负荷的感知，显示出三种自主反应模式，分别依赖于EDA、HF - HRV或这两种标记的组合。基于EDA、HRV和EDA/HRV的模式的存在可能解释了为什么以前主要采用单一生物信号的尝试在评估感知到的认知负荷时往往没有定论，从而证明了本方法在监测人类操作员与心理相关的工作量方面的附加价值。
