A semi-Markovian parametric model for slowly time-varying binary switchings
D. Wang · 1986
A method of modeling slowly time-varying binary switching elements with a periodic two-state semi-Markovian parametric model is presented. An optimal Maximum Likelihood (ML) estimator is derived for general holding time distributions. In addition, a slight restriction on the holding time distribution reduces the ML estimator into a Weighted Least Square (WLS) estimator. Finally, a linearization of the WLS equation results in a sub-optimal Linear Least Square (LLS) estimator which is non-iterative and involves finding the solution to a set of matrix equations, where the matrices are correlation matrices.