Classification of Human Postures

The classification of human postures is required in various aspects such as the description of human activities, the synthesis of humanoid robots’ movement, and in the research on robots cooperating with humans. The presented research aims to develop the method for posture classification using the data recorded by the RGB-D cameras. The method enables the recognition of the activity represented by the sequence of human postures viewed in the sagittal plane. The competitive neural network (cNN) was used. The method enables the fast recognition of human activity and is useful for ‘learning by observation’ robots.