Pose Estimation For Gesture Recovery In Occluded Television Videos