Geometric Brownian motion (GBM) random process model appears to be an excellent choice for modeling realizations of perclos signals

Introduction: The PERCLOS (PERcentage CLOSure) for one eye and a given time window of duration T seconds (typically 20) is the total time where the eye is closed more than, say, 80% with respect to some reference maximum opening, divided by the window length. Since PERCLOS is recognized as an important indicator of drowsiness, several drowsiness monitoring systems, e.g. for drivers, use it as their unique or primary parameter. All these systems compute PERCLOS "now" based upon the ("past") data in the window, so that an alarm may come too late to prevent an accident.

http://ift.tt/2GDj7KM