Towards a unified framework for hand-based methods in First Person Vision