The Research Of Nonlinear Time Series Based On Particle Filter Optimal Estimate

In this paper, nonlinear time series modeling based on the analysis of particle filter state transition of the optimal estimation. Time series modeling used the nonlinear and non-Gaussian particle filtering technique to and forecasting research.It is difficult to establish a real-time data trend analysis model for the predicted data object when a dynamic transfer process engender a structural mutation. In this paper, the particle filter estimated parameters of nonlinear time series model. First, the preliminary modeling of time series disturbanced the parameters as particle. Observations made the state transition with these models and parameters. State observations obtained the posterior probability density. The true value of the deviation and observations gained the weight of particle. From above metheds, state transition model and parameters obtained the best approximation.Real-time modeling of nonlinear time series dynamically constructured forecasting state transition model. Monte carlo simulation of particle filter used the state space in a series of weighted random sample. State transition set to approximate the system state probability density function of the posterior. The experiments show that the nonlinear time series based on particle filter optimal estimate algorithm has smaller mean square error than the particle filter. In the forecast, it is more accurate than traditional time series modeling.In this paper,the nonlinear time series based on particle filter optimal estimate and prediction method is applied to target tracking. The object template is used to a target observations of particle filter. State value is the image coordinates of pixel information from particle filter state transition equation. Real-time adaptive nonlinear time series modeling mined the regularity of target motion trajectory. Experiments show that the algorithm have some advantages as tracking performance. Nonlinear time series based on particle filter optimal estimate possessed faster tracking speed and higher precision.