Parameterisation of a stochastic model for human face identification

Abstract
Recent work on face identification using continuousdensity Hidden Markov Models (HMMs) has shownthat stochastic modelling can be used successfully toencode feature information. When frontal images offaces are sampled using top-bottom scanning, thereis a natural order in which the features appear andthis can be conveniently modelled using a top-bottomHMM. However, a top-bottom HMM is characterisedby different parameters, the choice of which has so farbeen based on subjective intuition. ...

This publication has 6 references indexed in Scilit: