Human Body Action Recognition System Using ELM Algorithm

Abstract: Human action recognition is the field of research that has gained importance in the recent years because of its wide range of applications spread across different fields. Human action recognition plays a crucial role in human-to- human interaction and also interpersonal relations as it provides the information about the identity, personality and psychological state of the person. The human ability to recognize other person’s activities is one of the major subjects of study of the scientific areas of computer vision and machine learning and because of this research many applications such as video surveillance systems, human-computer interaction, control free gaming systems, etc. require multiple action recognition system. Human action recognition also has a very significant use in security systems installed in public places to track any suspicious activity or threat. This project aims to develop an algorithm which can recognize actions such as jogging, bending, bowling, jumping, kicking, running etc. from the input video sequences. The multiple actions in the video sequence will be detected with each passing frame of the video sequence.

Keywords: Human action recognition, computer vision and machine learning.