An Optical Flow-based Action Recognition Algorithm

Abstract — This paper proposes a new technique for motion-based representation on the basis of optical flow analysis and random sample consensus (RANSAC) method. The method is based on the fact that an action can be characterized by the frequent movement of the optical flow points or interest points at different areas of the human figure. A combination of optical flow and RANSAC to identify a human body within a frame and the average percentage of change of interest points at various positions around the body are used as features for a particular action. Using the extracted features, a distance-based similarity measure and a support vector machine (SVM)-based classification technique have been exploited for recognition. From experimentations upon a standard motion database, it is found that the proposed method offers a very high degree of accuracy. I