A Bayesian approach for the estimation of AR coefficients from noisy biomedical data

In this paper we study the identification of AR parameters in a biomedical signal corrupted by additive white gaussian noise. The identification of AR parameter is treated as a signal estimation problem, whose aim is to obtain an estimate of the clean signal, given the noisy observations, and after that to obtain the noise free AR parameters. The novelty of our approach is the simultaneous estimation of AR parameter and the model order of the AR process. This is done adopting a Bayesian framework and using a special form for the prior of AR parameters. To obtain the solution we use the Variational Bayesian (VB) Framework. Simulation results have shown that the proposed approach correctly identifies the model order of AR process while at the same time produces an estimate for the AR parameters.