Accurate Vigilance Detection During Gait by Using Movement Artifact Removal

Human vigilance is a cognitive function that requires sustained attention toward change in the environment. Human vigilance detection is a widely investigated topic which can be accomplished by various approaches. Most studies have focused on stationary vigilance detection due to the high effect of interference such as motion artifacts which are prominent in common movements such as walking. Functional Near-Infrared Spectroscopy is a preferred modality in vigilance detection due to the safe nature, the low cost and ease of implementation. fNIRS is not immune to motion artifact interference, and therefore human vigilance detection performance would be severely degraded when studied during locomotion. Properly treating and removing walking-induced motion artifacts from the contaminated signals is crucial to ensure accurate vigilance detection. This study compared the vigilance level detection during both stationary and walking states and confirmed that the performance of vigilance level detection during walking is significantly deteriorated (with a p<; 0.05). Further, this study explored motion artifact removal and applied machine learning methods. Results reveal the vigilance detection during walking has a comparable performance to the stationary state when the motion artifacts are estimated and removed.