| The problem of estimating the state of a stochastic dynamical system from noisy observations taken on the state is of central importance in engineering. Noise filtering is an important part of processing a real signal sequence. The more common approach in modeling the noise is to assume that it has Gaussian distribution in addition to some low percentage of additive impulsive noise. The filtering involves the removal or reduction of impulsive noise along with the Gaussian noise while keeping the original signal.; In this thesis, we present a fuzzy filter, which is based on fuzzy logic theory and Kalman filter. It is similar to Kalman filter when a linear system with Gaussian noise is considered. However, when non-Gaussian noise is introduced, it is shown that fuzzy filter is outperforming Kalman filter, while Kalman filter does not work well. We demonstrate the performances of Kalman filter and fuzzy filter for position estimation application under different kinds of circumstances. The comparisons are made to draw conclusions. |
