In this paper we present a Monte Carlo EM algorithm
for learning the parameters of a state-space model with a Markov switching.
Since the expectations in the E step are intractable, we consider an
implementation based on the Gibbs sample. The rate of convergence is improved
using a nesting algorithm and Rao-Blackwellised forms. We illustrate the
performance of the proposed method for simulated and experimental physiological
data.