Parameter estimation of nonhomogeneous hidden semi-Markov models

The article is devoted to the study of nonhomogeneous hidden semi-Markov models, and proposes a methodology for estimating parameters in the case where the change of state sojourn time distributions and transition probability matrices is time-limited. The obtained results allow the model to be applied to the description of systems with time-varying structure. The proposed methodology can be used in time series analysis tasks with structural changes, as well as in optimal control problems for stochastic systems where it is important to account for changes in the process behavior over time.