Model-level data-driven sub-units for signs in videos of continuous Sign Language | IEEE Conference Publication | IEEE Xplore