earticle

영어

This paper proposes a method to analyze human actions and provide alarm service through a smart-phone. The proposed method extracts objects using camera input images and modeling background. This paper defines a action recognition parameter as an extracted object’ histogram and its joints’ information about human action recognition and analyzes changes in defined parameter to determine conditions for recognizing human actions. Furthermore, in order to systemically classify actions, the proposed method designs a decision tree and combines the defined conditions with a conditional expression of tree node, which enables effective action recognition and classification. Recognized actions are transferred to a smart phone as a simple value and the transferred result is displayed on the phone. The proposed method is tested from a video of a webcam about nine actions of four persons. As for action recognition and alarm service of a smart phone, the proposed method works well and effectively.