Recognition of stochastic signals using Markov chains of order s

For discriminant analysis of a stochastic signal, we propose to change the signal into a discrete signal described by a Markov chain of order s; the problem of the initial signal recognition is considered in a space of the Markov chain parameters. The Bayesian decision rule and plug-in Bayesian decision rule for Markov chain classification are constructed. The misclassification probability of the decision rules under an asymptotic of contiguous classes is found. The proposed approach is successfully applied for diagnostics of the coronary ischemic disease by recognition of biomedicine signals.