Characterization of wake and stage1 sleep architecture is done using multidimensional Spectral entropy (SE) feature ranking.
A wake/stage1sleep scoring model is proposed using SE computed per second unlike 30 sec interval used in manual scoring.
The classifier models are built on a robust training method that randomly selects 60% of the dataset for training.
The proposed methods show F4 region with high discrimination between groups linked to reduced activity in stage1 as reported in medical literature.