Skull recognition based on Embedded Hidden Markov models

Skull recognition system based on embedded hidden Markov model (EHMM) has been developed. The coefficients of two dimensional discrete cosine transform (2DDCT) are used to construct an efficient set of observation vectors. Different from previous systems that were only capable of operating on frontal skull images, the EHMM method presented here, making use of the information of skull shape and holistic structure, reflects the relation of adjacent organs and reduces the sensitivity to the skull deflection. Experiments are carried out on a database of 480 different skull images, which consists of 12 different head orientations, performed by 40 individuals. The recognition accuracy is 93.3%, which proves that this method is robust against the variation of skull orientation.