Human activity recognition based on Gaussian mixture model and directive local binary pattern

Recognizing human activities has become an active research area in computer vision because of its promising need and use in many applications. In this paper, we represent a method for recognizing human activities in video. In order to recognize different activities accurately a distinctive and informative feature vector is required which can encode more distinctive information. We introduce a feature named Directive Local Binary Pattern (DLBP) which we develop on the basis of our experimental result that is orientation information is more informative than magnitude information for binary silhouette based recognition. The proposed Directive Local Binary Pattern (DLBP) incorporates orientation information with intensity differences of binary silhouette images. It is further combined with Edge Orientation Histogram (EOH) and forms a more distinctive and informative feature set. By means of the extracted features, Support vector machine is used for training and recognition which is a robust classifier. The proposed method for recognizing human activities is experimented on different videos containing various moving humans and the outcomes of our method are encouraging.