Using photoplethysmography based features as indicators of drowsiness: Preliminary results

Abstract

Drowsiness, a transitional psychological state between alertness and sleepiness, is one of the leading reasons for the increased risk of accidents, particularly in commercial aviation. Though electroencephalography (EEG) is widely used for drowsiness assessment, it is impractical for unobtrusive airborne monitoring due to discomfort induced by the number of electrodes touching the scalp. In this work, we investigated the potential of pulse arrival time (PAT), and photoplethysmogram (PPG) features as indicators of drowsiness. Simultaneous vertical electrooculogram (EOG), ear PPG, and electrocardiogram (ECG) were recorded from 18 commercially-rated pilots from 02:00 AM to 04:30 AM. ECG R-peaks and PPG peaks were detected and used for calculation of PAT and heart rate to observe their changes during drowsiness. During the drowsy periods, the PAT and RR intervals increased (i.e., decrease in heart rate) significantly (p<0.05) compared to baseline. Features based on only PPG also showed a significant increase during drowsiness. However, decrease in PAT/RR during drowsiness for most of the subjects indicates that increase in PAT is not linearly correlated to RR interval and might be reflecting changes in arterial stiffness. Hence, PAT and derived PPG based features with continuous heart rate monitoring can serve as useful indicators of early drowsiness detection.