Unsupervised classification of extreme facial events using active appearance models tracking for sign language videos | IEEE Conference Publication | IEEE Xplore