Classification approach to determining emotional state of person based on analysis of video information

Abstract

This work considers recognition of person’s emotional state with body movements and postures as a problem of group classification of multi-attribute objects. The analysis of existing approaches to the recognition of the human emotional state based on video information was made. An approach proposed includes using markerless recognition of body movements based on video information, identification of poses and movements with a model of human motor activity, characteristics description of postures and movements in a limited natural language and the group classification method of poses and movements having verbal description for emotional state identification.