Multiclass SVM Based Real-Time Hand Gesture Recognition

Abstract: For the communication between human being and in sign language hand gestures plays an important role. Deaf and deam people survive their life communication with hand gestures. Experimental scheme of the course of action uses solid position economical web camera mutually 10 mega pixel resolution mounted on the outstrip of monitor of computer which captures snapshot by Red Green Blue [RGB] enlarge space from tense distance The work is isolated into four stages one as image preprocessing, region extraction, feature extraction, feature matching. First it take the continuous images from the web camera mounted on the top of the machine. At the next level converts captured RGB conception into gray threshold approach with noise removed by the agency of median filter and Guassian filter, followed by morphological operations. At the third stage the features are extracted using HOG and classiffed using SVM algorithm. The paper include some example of dynamic hand gesture recognition related to some actions. Traning dataset consist of 20 samples of differ symbols.

Keywords: Image Preprocessing; Region Extraction; Feature Extraction; Median Filter; Support Vector Machine.