Visual analysis framework for two-person interaction
Thi Thi Zin, Junpei Kurohane · 2015
In this paper, a novel approach to two person interaction method is presented in which pose representation is based on the feature of silhouette images. Today human activity recognition and analysis has a tremendous potential to impact a wide range of applications from surveillance to human computer interfaces to content based video retrieval. Specifically, the proposed method makes use of human silhouettes to classify actions and interactions of human present in a scene video. The classes of interactions will include punching, pushing, kicking, hand-shaking, and hugging. Moreover, the detected interactions are further divided into violence or non-violence so that a suitable security measures would be taken. To confirm the validity of the proposed method, the experimental results are carried out by using the publicly available dataset.