Human posture classification using skeleton information

Behavior understanding in images and sequence of images is a widely researched topic these days. One important step towards behavior understanding is Human Posture Recognition that can be used for recognizing activities occurring in a particular scene. In this research, an algorithm for posture recognition in still images using 2D pose information from human skeleton was implemented. An approach based on angles and distances between joints was adopted for classification between sitting and standing postures. The results showed that desired postures were recognized with a good average accuracy of 95%.