Human Action Recognition in Video Under Clutter and Moving Background

Abstract

Behavior or action recognition in video sequences has becoming a more interesting and active research for computer vision. Applications of human action recognition, such as video retrieval, video surveillance and video event analysis are expanded extensively. Generally the video data used are based on fixed camera with stationary background, well-controlled environment and under simple actions. However, in real human action cases, the actions of human behavior are often complex and the background are cluttered with illumination changes, different human body size and moving camera. These make the real video much more complex. In this study, study to improve accuracies of human action recognition under clutter and moving background is proposed. The recognition scheme is based on the space-time interest point (STIP) and naïve Bayes based mutual information maximization (NBMIM). Methods including the selection of robust feature points based on camera motion estimation, analysis and correlations of the important STIP features during the training stage and weighting mechanism for action recognition to improve the recognition rate are used. Experimental results using the YouTube dataset indicate the effectiveness of the proposed scheme.