On Evidential Markov Chains

Hélène Soubaras.

Evidential Markov chains (EMCs) are a generalization of classical Markov chains to the Dempster-Shafer theory, replacing the involved states by sets of states. They have been proposed recently in the particular field of an image segmentation application, as hidden models. With the aim to propose them as a more general tool, this paper explores new theoretical aspects about the conditioning of belief functions and the comparison to classical Markov chains and HMMs will be discussed. New computation tools based on matrices are proposed. The potential application domains seem promising in the information-based decision-support systems and an example is given.

PDF full paper