Gibbs random sequences Bayes estimation based on stochastic relaxation

Abstract

An opportunity of Metropolis-Hastings stochastic relaxation procedure application for optimal Gibbs random message interpolation is investigated. A conditional-Gaussian random sequence as an example of Gibbs sequence observed in white Gaussian noise is used as message model. Such models are used, e.g. for speech signal description. The basic Metropolis-Hastings algorithm is described and some experimental results are presented.