Fixed parameter estimation method using gaussian particle filter

• Authors: Lixin Wang
• Venue: ICIC'06 Proceedings of the 2006 international conference on Computational Intelligence and Bioinformatics - Volume Part III
• Year: 2006

The degeneracy problems of general particle filtering frequently occur. Although this kind of problems can be mitigated by resampling, but the diversity characteristic between particles may be lost because the higher weighted particles will be replicated and the lower weighted particles will be discarded. For parameter-fixed application cases, the standard particle filter is invalid as no importance density function can be sampled for new particles required by the predictive distribution, and particles will quickly be exhausted. This paper proposes a new method for the parameter-fixed estimation by use of Gaussian particle filter, which can avoid making particles exhausted and can improve the estimation performance. Refer to a practical example of Direction of Arrived (DOA) estimation for coherent signals propagated in space with multi-path fading, the computer simulation has been performed. The simulation results have indicated that the performance of the new method is rather than general particle filtering.