Human action recognition using a corners and blob detector with different classification methods

Abstract

Human Action Recognition trying to recognize the motion and movement of the person, the recognition of human movement is a very important field in many applications of security such as for smart surveillance and monitoring systems, also in another application such as computer interaction and entertainment. This paper tries to detect and identify the action of human but there is a big challenge in recognize similar actions for example walking and running. The dataset has been used is KTH dataset and the methods are harrier detector and blob detector to extract the corners from each frame. the second topic has been discuses in this paper is to examine between two different classification methods: supervised (SVM) and unsupervised (KNN) Machine Learning to find the most suitable way classification for this paper, we were find that the KNN classification is overcome the SVM Classification in this work. After the result has been extracted we find the using of KNN method is the most suitable for this topic. In future work we hope using more features to extract the corner to increase the efficiency of classification process.