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.