Exploiting Phonological Constraints For Handshape Recognition In Sign Language Video