A Robust Framework for Effective Human Activity Analysis

Human activity analysis is an interesting and challenging problem among the researchers of computer vision area. The applications of human activity analysis are monitoring and surveillance. There are various surveillance approaches available in the literature for witnessing activities, events, or persons. In this paper, we present a robust framework for human action analysis. In the proposed framework, we extract the features named as generate motion image from frames deviation. Random forest is used as a feature classifier. To show the robustness of the proposed framework, we analyze and classify the publicly available HMDB dataset. The average accuracy of classification is 46.83% achieved.