In this paper a comparison between two popular feature extraction methods is presented. Scale-invariant feature transform (or SIFT) is the first method. The Speeded up robust features (or SURF) is presented as second. These two methods are tested on set of depth maps. Ten defined gestures of left hand are in these depth maps. The Microsoft Kinect camera is used for capturing the images . The Support vector machine (or SVM) is used as classification method. The results are accuracy of SVM prediction on selected images.