A mode preserving particle filter

A new particle filtering technique is proposed which addresses the problem of mode preservation in the approximation of multi-modal posteriors. The key is a resampling step which attempts to distribute samples evenly over the state space. This idea is shown to produce enormous benefits compared to the optimal importance density in a test problem requiring the retention of several modes for an extended period.

• Type: Journal
• Year: 2011
• Where: ICASSP
• Authors: Mark R. Morelande, Alan M. Zhang
• Keywords: ICASSP 2011, Multi-modal Posteriors, Optimal Importance Density, Particle filtering Technique, Signal Processing