Detection of multiple, co-occurring actions in video frames

Abstract

We propose a systematic frame work for the automatic detection of multiple human actions within the same frame in realistic and diverse video settings. One of the major challenges is the process of recognizing and understanding of human actions from videos with large variations resulting from camera motions, changes in human appearance, pose changes, scale changes and back ground clutter etc. In this paper 8 different human actions, occurring independently as well as co-occurring with one of the other seven, are detected. The detected human actions are classified based on the clustering of spatio-temporal features and classified using the multi-channel nonlinear SVM. Applying this method to publicly available data has shown promising results.