A Unified System for Segmentation and Tracking of Face and Hands in Sign Language Recognition