Finger Spelling Recognition for Nepali Sign Language

The use of Human-Computer Interaction (HCI) has improved day by day. Hand gesture recognition system [1-3] can be used to interface computer with humans using hand gesture. It is applicable in the areas like virtual environment, smart surveillance, sign language translation, medical system, robot control, etc [4, 5]. The paper elaborated the mechanism to identify the Nepali Sign Language with the help of hand gesture using shape information. It uses radial approach for dividing the segmented image and to obtain the sampled points making the process scaling invariant. The feature set is obtained by using the Freeman chain code and Vertex Chain Code (VCC) techniques. Skin color model has been used to identify the hand from simple background. Further processing is done in order to remove the unwanted noise and areas. Blob analysis is then carried out in order to extract the hand gesture from the image considering the largest blob in the image. The centroid of the image is identified, and the image is divided equally by plotting a line at certain angle from the centroid of the image. For sampling of the image, the point of intersection of line and the boundary of image is taken. For the classification purpose, k-NN classification technique is used. Confusion matrix approach is used to authenticate its accuracy. The accuracy using sampling for radial approach and use of Freeman chain code for feature extraction was found to be efficient than sampling using radial approach and use of transcribing vertex chain code technique and sampling using grid and use of Freeman chain code for feature extraction.