Motion Clustering-based Action Recognition Technique Using Optical Flow

A new technique for action clustering-based human action representation on the basis of optical flow analysis and random sample consensus (RANSAC) method is proposed in this paper. The apparent motion of the human subject with respect to the background is detected using optical flow analysis, while the RANSAC algorithm is used to filter out unwanted interested points. From the remaining key interest points, the human subject is localized and the rectangular area surrounding the human body is segmented both horizontally and vertically. Next, the percentage of change of interest points at every small blocks at the intersections of horizontal and vertical segments from frame to frame are accumulated in matrix form for different persons performing the same action. An average of all these matrices is used as a feature vector for that particular action. In addition, the change in the position of the person along X-axis and Y-axis are cumulated for an action and included in the feature vectors. For the purpose of recognition using the extracted feature vectors, a distance-based similarity measure and a support vector machine (SVM)-based classifiers have been exploited. From extensive experimentations upon benchmark motion databases, it is found that the proposed method offers not only a very high degree of accuracy but also computational savings.