Estimation of sleepiness using frequency components of pupillary response

Frequency components of pupil responses were extracted from power spectrum density of pupillary change during eye sleepiness tests to conduct an efficient assessment of sleepiness using pupil response. These factor components were applied to two-class classifications of subjective sleepiness using discriminate analysis. The correct rates of the classification during 5 minutes observation were 71% using the Stanford Sleepiness Score (SSS) and a Visual Analogue Scale (VAS), and 94% using the Epworth sleepiness score. Differences in contribution ratios for factor components were shown between the two kinds of subjective sleepiness.