Face pose estimation based on EHMM and SVM

Estimating pose of human face plays important role in lip reading. Mouth shape varies on face pose variation. Therefore, pose estimation will be front end of lip reading system. Embedded HMM can be successfully applied in estimating pose of human face. Advantage of EHMM based pose estimation method is non-sensitive in parallel shift and scale varying. On the other side, SVM is the classifier which has strong discrimination ability. This paper suggests EHMM and SVM based human face pose estimation method. It was experimented on CubiC FacePix database and was estimated with several range of angle. Experiment proved validity of this algorithm.