A Review of EEG-based Prediction of Microsleep

Abstract

Microsleep is a brief, complete, and unintentional sleep-related loss of consciousness. The duration of microsleeps and
the probability of accidents are highly correlated. In extended monotonous activities, the consequences of microsleeps
are therefore often catastrophic. Electroencephalogram (EEG) signals have widely been used for the early detection
of microsleeps. This paper comprehensively reviews several EEG-based microsleep detection/prediction techniques
published in academic conferences and journal articles published between January 2005 and January 2023. The review
specifically discusses the preprocessing, feature extraction, and feature selection techniques collectively known as
feature engineering, machine learning algorithms, and overall performance metrics. The review finally presents future
directions that could help improve the overall EEG-based microsleep prediction systems