The particle filter is usually used as a tracking algorithm in non-linear under the Bayesian tracking framework. However, the problems of degeneracy and impoverishment degrade its performance. The particle filter is thereafter enhanced by evolutionary optimization, in particular, Particle Swarm Optimization (PSO) is used in this thesis due to its capability of optimizing non-linear problems. In this thesis, the PSO enhanced particle filter is reviewed followed by an analysis of its drawbacks. Then, a novel sampling mechanism for the particle filter is proposed. This method generates particles via the PSO process and estimates the importance distribution from all the particles generated. This ensures that particles are located in high likelihood regions while still maintaining a certain level of diversity. This sampling mechanism is then used together with the marginal particle filter. The proposed method’s superiority in performance over the conventional particle filter is then demonstrated by simulations.