Détection bayésienne de saut de fréquence par des méthodes MCMC

The problem addressed in this paper is the off-line detection and estimation of frequency jump in a noisy complex sinusoid. A Bayesian approach using Markov chain Monte Carlo methods for estimating the change time r and the frequencies before and after the jump only with observations is proposed. A adaptive version of the random walk Metropolis-Hastings (M-H) algorithm is used allowing to adjust automatically the algorithm parameters. Performances of detection estimated by the distribution of change time estimation error and ones of the estimation of change amplitude are presented.