Research And Application Of An Improved Particle Filter Algorithm

With the development of statistical signal processing technology, nonlinear filtering technique attracts more and more attention. Its basic task is to recursively evaluate the states which can not be observed. Particles filter which achieves Bayesian filtering by using non-parametric Monte Carlo simulation methods was developed recent years as a nonlinear filtering. The state space is described by particles, and the posterior distribution of the state was approached by using particles and the discrete random measures of the weights. But after a few iterations steps, particle filter has the degradation problems. In this dissertation for dealing with the degradation problems, an improved particles filter algorithm was proposed.Firstly, the basic idea of improving the particles filter algorithm is presented. The U-particles filter algorithm with the MCMC step is based on the combination of U-particles filter and MCMC particles filter. The latest measurements are introduced to state estimation by UKF in the algorithm to obtain a better importance distribution, and then the MCMC method is used after the re-sampling step to increase the diversity of particles for improving the performance.Secondly, the performance of U particle filter with the MCMC step is analyzed, the algorithm is improved from samples and the use of MCMC, for the purpose of reducing the cost and real increasing speed of computation speed.Finally, the basic particle filter algorithm and improved algorithm proposed in this dissertation is applied to the blind channel equalization experiment. The bit error rate of the particle filter algorithm and EKF algorithm are compared, and then the bit error rate of a variety of existing particle filter algorithm are compared. lastly the bit error rate of the existing algorithm which has the lowest bit error rate and the proposed algorithm are compared,and the root mean square error of all kinds of algorithms is analyzed.Both performance analysis and simulation results showed that the improved algorithm can reduce RMSE and BER, and at the same time it can increase the computional speed. The application of blind equalization in channel by the improved algorithm meets the demand of accuracy and real-time in communication.