Hand gesture recognition based on depth map

In this paper a proposed method for gesture recognition using depth map image is presented. Three different feature extraction methods are presented too. All of them are based on Radon transform. First method reduces the feature vector of Radon transform by averaging its values. Next approach uses discrete cosine transform on radon graph, thus the energy of graph is reorganized. This approach concentrates the energy of graph to upper left corner and the rest can be cutoff with minimal error. Last approach applies R transform on radon graph, those create 1D shape-curve in specific direction. For the classification of features vectors, the support vector machine is used. Finally, the experimental results of all descriptors are shown.