State Space Modelling Using Particle Filtering

In any signal/image/video/process engineering state modelling takes leading role. In state modelling, On-line state estimation is a key element. When state functions are highly non-linear and the posterior probability of the state is non-Gaussian, then any one of the conventional filter could not provide efficient and satisfactory results. It is typically crucial to process data online as it arrives, it is both from the point of view of storage cost as well as for rapid adaptation to changing signal characteristics. This paper proposes an alternative approach of conventional filtering techniques is particle filtering. We reviewed Bayesian algorithm for non-linear and non-Gaussian tracking problems with a focus on particle filtering. Particle filters are working based on the SEQUENTIAL MONTE-CARLO METHOD’s. Sequential* Monte-Carlo method based on point mass representations of probability densities, which can be mostly applying to state space modelling and generalises the traditional Kalman filtering methods. Particle filtering is described prior to discussing a number of implementation issues used for the estimation task. Furthermore, MARKOV CHAIN MONTE-CARLO METHOD is proposed to enhance particle filters where the estimation of the initial conditions is poor. The effectiveness of particle filtering of state estimation is explained.