AN ALGORITHM FOR DETECTING NON-VERBAL MARKERS OF HUMAN BEHAVIOR ON VIDEO

Abstract

Microexpressions are unconscious, short-term non-verbal signals that allow to determine the emotional state of a person. Microexpressions occur when a person blocks emotions or hides true intentions. Determining non-verbal signals becomes an urgent task in situations where lying or hiding information leads to resource or financial losses, affects the safety and health of other people. The spread of online conferences opens up the possibility of programmatic processing of a human speech video channel to analyze emotions and behavior in order to identify the congruence or inconsistency of person's statements. The article discusses computer vision and machine learning methods that allow extracting and analyzing person's face from a video channel to determine its non-verbal markers and emotional state. The method of facial landmarks, key points of the face, classification of human emotions by facial landmarks, detection of blinking and turning of a person during speech are considered in detail.