The complexity of the problem to evaluate the human functional state and fatigue can be divided into the research of pathological and physiological forms of fatigue where current research is focused on the physiological form and characterized as proportional to the activity and acute as the functional state can be recovered by sleep and rest. Sleepiness or drowsiness is among the key indications of physiological fatigue and persists throughout the human daily cycle. The risks often associated with a certain gradation of sleepiness, classification of which is described in this article are mainly considered in the context of determining the degree of sleepiness of the driver or operation safety in the workplace. The main methods to be addressed in this paper are the selection of a set of sleepiness characteristics to determine the degree of sleepiness, the selection of the most informative traits for decision-making algorithms, and the resolution of problems associated with visual information extraction from the video. A methodology with multiple algorithms is proposed to obtain values for sleepiness characteristics from a video with human face regions. A total of 28 characteristics were obtained. The decision-making algorithms were trained by using a video database that provided sleep values for the existing video and validated with volunteer video dataset. The classification space three degrees of sleepiness was applied to label the videos, and two classes where low levels of alertness and severe sleepiness were combined into one class. For the prediction of the three degrees of sleepiness, the accuracy of decision algorithms ranged from 37 % to 53 %, which in all cases was better than the free guess limit of 33 %. For the prediction of two levels of sleepiness, the accuracy of decision algorithms ranged from 67 % to 73 %, which in all cases was better than the 50 % limit of free guess.