Modelling of the Markovian behaviour of mobile satellite channels with MCMC methods

Classical models for mobile satellite channels are based on Markov chains where each state is linked to special propagation conditions. In this paper, we focus on the methodology of extraction of hidden Markov model (HMM) from experimental data to describe the time fluctuations of received power in a mobile satellite service (MSS) context. The method developed in this paper is based on a MCMC (Monte Carlo Markov Chain) training phase associated with a k-means classification. Its complexity is reduced when compared to traditional MCMC method. Contrary to existing detection methods, nearly no prior are necessary and it enables an accurate estimation of the HMM parameters and the possibility to detect the evolution of the channel sequentially. Furthermore, a sequential method, based on estimated HMM parameters, is presented and enables to follow the on-line variation of the state sequence.