Sign Language Interpretation Using Hand Gesture Recognition

Abstract

Hand gesture recognition is an active area of research in computer vision with wide range of application in the area of sign language recognition, games, human computer interaction etc. Sign language is the basic communication method with speech and hearing imapaired. The aim of this thesis work is to develop a system which will translate hand gesture into the corresponding sign language entity. The various stages of the system are: segmentation, feature extraction and classification.

The segmentation method is used to locate hands in the images. Gaussian model is used to find the skin-color distribution. This is further used to obtain likelihood of skin for any pixel of image. These images are transformed into binary image by proper thresholding. The binary image of the hand gesture is processed with erosion to eliminate the noise and then processed with dilation to recover the original shape.

Three methods for recognition and classification of hand gesture are presented boundary contour method, Medial axis (Skeletonization) method and circle based method. Common gesture of fingers to indicate the numerals have been used for testing. The results indicate that it is possible to identify the number from the gesture of fingers.