Non-Intrusive Cognitive Load Estimation Using Footstep-Induced Structural Vibration Signals

Abstract

Cognitive load estimation is crucial for understanding human performance during demanding tasks. Elevated cognitive load can impair decision-making and increase the risk of fatigue or burnout. Consequently, real-time monitoring of cognitive load is essential in high-stakes environments such as healthcare, where even subtle changes in human movement can have significant consequences. Conventional methods for cognitive load assessment predominantly rely on neuro-imaging, physiological signals, or subjective reporting, which are often not real-time, intrusive and can cause user discomfort. In this study, we propose the use of structural vibrations as a non-intrusive and privacy-preserving modality for cognitive load estimation, offering a promising alternative in healthcare and other sensitive domains. We conducted controlled experiments with eight participants under two walking conditions: a baseline of normal walking and a dual-task condition where participants walked while engaging in cognitively demanding interactions with a conversational agent. Comparative analysis revealed distinct changes in gait dynamics and footstep-induced structural vibration patterns under cognitive load. Our experimental results identified clear sensor signatures corresponding to varying levels of cognitive demand, demonstrating the potential of structural vibration metrics as indicators of cognitive load. We achieved 68.00% accuracy in cognitive load classification using structural vibration signals. Furthermore, for person identification, we attained an accuracy of 82.00% based on structural vibration data.