Sign Language Gesture Recognition using Zernike Moments and DTW

Since the last few decades, a dominant area of research in the vision community has been the gesture recognition, mainly for the purpose of Human Computer Interaction (HCI) and recognition of sign language. In this paper, we are using Zernike Moments as shape descriptors. The proposed system for recognizing sign language mainly consists of following five modules: (1) gesture segmentation based on motion detection analysis, (2) real time detection of both hand regions and face region, (3) key frame extraction for removing redundant frames, (4) the feature extraction phase consists of tracking the hands trajectory in terms of orientation, tracking distance of hands from the centre of the face and determining the hand posture using rotation invariant Zernike Moments and finally (5) gesture recognition based on these extracted features using Dynamic Time Warping (DTW) methodology.