Presentation
Real-time Vigilance Monitoring via Biosignals: A Scoping Review
SessionWednesday Poster Session
DescriptionThis study presents a scoping review of real-time vigilance monitoring systems using physiological biosignals, with a focus on their applicability in safety-critical domains. Vigilance, defined as the ability to sustain attention over prolonged periods, is essential for tasks such as driving, aviation, and industrial operations, where lapses can lead to significant risks. To evaluate current advancements, a systematic review following PRISMA guidelines was conducted across IEEE, ProQuest, PubMed, and ScienceDirect databases. After applying inclusion and exclusion criteria, 16 studies were selected for analysis.
The findings indicate that electroencephalography (EEG) is the most frequently used sensor modality, while heart rate variability (HRV), electrooculography (EOG), photoplethysmography (PPG), and galvanic skin response (GSR) are less commonly employed. Most studies rely on supervised machine learning models, particularly for binary classification of vigilance states such as alert versus drowsy. Applications are heavily concentrated in automotive contexts, with limited cross-domain validation.
Despite advances in sensors and computational modeling in the era of artificial intelligence, current approaches often emphasize threshold-based detection rather than continuous monitoring of vigilance dynamics. This review highlights the need for more robust multimodal integration, real-world validation, and adaptive modeling frameworks to enable reliable deployment of real-time vigilance monitoring systems across diverse operational environments.
The findings indicate that electroencephalography (EEG) is the most frequently used sensor modality, while heart rate variability (HRV), electrooculography (EOG), photoplethysmography (PPG), and galvanic skin response (GSR) are less commonly employed. Most studies rely on supervised machine learning models, particularly for binary classification of vigilance states such as alert versus drowsy. Applications are heavily concentrated in automotive contexts, with limited cross-domain validation.
Despite advances in sensors and computational modeling in the era of artificial intelligence, current approaches often emphasize threshold-based detection rather than continuous monitoring of vigilance dynamics. This review highlights the need for more robust multimodal integration, real-world validation, and adaptive modeling frameworks to enable reliable deployment of real-time vigilance monitoring systems across diverse operational environments.
Event Type
Poster
TimeWednesday, October 21st5:30pm - 6:30pm PDT
Location