Real-Time Cognitive Overload Monitoring in Simulated Navigation Using Wearable Biosensors

Cognitive overload poses a significant challenge to driver safety, especially in the context of information-rich navigation systems. How to effectively solve the cognitive load problem caused by information overload during the navigation process and achieve real-time monitoring and on-demand prompt of navigation information has become a difficult problem faced by intelligent navigation. In this study, wearable biosensor device EmotiBit was used to extract physiological signals, including electrodermal activity (EDA) and photoplethysmogram (PPG), from drivers in virtual navigation scenarios. We extracted effective features related to skin conductance and heart rate from these physiological signals and selected support vector machine (SVM), logistic regression (LR), random forest (RF), and extreme gradient boosting (XGBoost) models as classifiers to develop a real-time monitoring model for cognitive overload. The results reveal a remarkable accuracy rate of 88.75% achieved by the random forest model, surpassing the performance of the other three algorithms, which demonstrated the feasibility of detecting cognitive load in users during navigation processes through real-time monitoring of skin conductance activity parameters. Feature importance analysis uncovered that the second-order difference of electrodermal activity and its level emerged as the most influential factors for cognitive load detection. This study underscores the critical significance of real-time monitoring of cognitive load in drivers to optimize navigation services and elevate road safety standards. The findings pave the way for future advancements in utilizing wearable technology for cognitive state monitoring and offer valuable implications for improving navigation systems and enhancing user experience on the road.认知过载对驾驶员安全构成重大挑战，尤其在信息密集型导航系统的应用场景中更为突出。如何有效解决导航过程中信息过载引发的认知负荷问题，实现导航信息的实时监测与按需推送，已成为智能导航领域亟待攻克的核心难题。本研究采用可穿戴生物传感器设备EmotiBit，从虚拟导航场景中的驾驶员身上采集生理信号，包括皮肤电活动（electrodermal activity, EDA）与光电容积描记图（photoplethysmogram, PPG）。研究团队从上述生理信号中提取与皮肤电导、心率相关的有效特征，并选取支持向量机（support vector machine, SVM）、逻辑回归（logistic regression, LR）、随机森林（random forest, RF）以及极端梯度提升（extreme gradient boosting, XGBoost）作为分类器，构建认知过载实时监测模型。实验结果显示，随机森林模型的准确率可达88.75%，显著优于其余三种算法，证明了通过实时监测皮肤电导活动参数，实现导航过程中用户认知负荷检测的可行性。特征重要性分析表明，皮肤电活动的二阶差分及其水平值是影响认知过载检测的最关键因素。本研究凸显了实时监测驾驶员认知负荷，对于优化导航服务、提升道路安全标准的重要意义。本研究成果为未来利用可穿戴技术实现认知状态监测的发展奠定了基础，并为优化导航系统、提升道路用户体验提供了重要参考价值。