Sign Gesture Recongnition Using Support Vector Machine

A tool for recognizing alphabet level continuous American Sign Language using Support Vector Machine to track the sign languages represented with hands is presented. Six letters are trained and recognized and got an efficiency of 92.13%. The static images of hand gestures representing the six letters are taken in a camera and processed for training and recognition. The image taken in the camera was so large to process so we resized the image to one eighth of its original size. Then the image is converted to gray and the edges of it are found out using the Sobel filter. Since our point of interest is the gesture made with hand we find out the largest three among all the connected components which would give another image as the output having only the boundary of the sign leaving behind the rest of the objects present in the image which are unnecessary. The coordinates of the edges are given as the input to the Support Vector Machine which will train and classify the same so that that next time when a test data is given it would get classified accordingly. Both the row and column position of the final image (consist of only three connected components) are taken as the elements of sample matrix used for training the images. The âSupport Vector Machine´ tool is used for classification and training.