Wayfinding Information Cognitive Load Classification Based on Functional Near-Infrared Spectroscopy

Abstract

Amid the rapid development of building information technologies, wayfinding information has become more accessible to building users and first responders. As a result, a realistic risk of cognitive load related to the wayfinding information processing starts to emerge. As cognition-driven adaptive wayfinding information systems become increasingly used to overcome challenges of cognition overload due to overwhelming information, a practical and noninvasive method to monitor and classify cognitive loads during the processing of wayfinding information is needed. This paper tested a functional near-infrared spectroscopy (fNIRS)-based method to identify cognitive load related to wayfinding information processing. The method provides a holistic fNIRS signal analytical pipeline to extract hemodynamic response features in the prefrontal cortex (PFC) for cognitive load classification. A human-subject experiment (N=15) based on the Sternberg working memory test was performed to model the relationship between fNIRS features and cognitive load. Personalized models were evaluated to capture individual differences and identify unique contributing features to each person. The results showed that the fNIRS-based model can help classify cognitive load changes driven by the different levels of task difficulty with satisfactory performance (avgerage accuracy rate 70.02%±4.41%). The findings also demonstrated that personalized models, instead of universal models, are needed for classifying cognitive load based on neuroimaging data. fNIRS has considerable advantages over other neuroimaging methods in cognitive load classification given its robustness to motion artifacts and the satisfactory predictability.