Pose estimation and tracking for sign language recognition

Abstract

In this work we combine several state-of-the-art algorithms into a single pose tracking pipeline. We recognize motion, relative position and pose of human hands and face as the main information carriers as we separate the problem into three main tasks: hand and face detection, hand and face pose estimation and tracking hands and face in dynamic environment. We also propose to use measurements from pose estimation steps as a visual information indicator to solve dynamic tracking assignment problem.