Extended Kalman filter and extended particle Kalman filter for non linear estimation problems

This article presents a methodology based on algorithms: filter extended Kalman and particle filter extended Kalman to study the problem of parameter estimation in dynamic models with non-linear structures, but with errors of observation Gaussian, arises in the form of State space model, where the no-observed-system States are treated as parametersrecursive Bayesian inference techniques are used to predict and update distribution to rear and marginal States. We illustrate the proposal estimating and reconstructing States of Henon maps and Lorenz, was also reconstructed States and morphological properties of a synthetic model of an electrocardiogram signals. The results show that filters have good behavior in the estimation of the States. Finally it evaluates the behavior of algorithms in terms of the empirical standard deviation and the times of computer CPU, with small variations in the estimated errors and rapid execution in real time.