An improved particle filter algorithm based on neural network for visual tracking

Abstract

Due to the shortcoming of constructing importance density in general particle filter, we propose an improved algorithm based on neural network to optimize the choice of importance density. It is proved to be more efficient than the general algorithm in the same sample size. This algorithm adjusts the samples drawn from prior density with general regression neural network (GRNN), and makes them approximate the importance density. Finally, the new algorithm is used to solve the target-tracking problem. Simulation shows that the proposed algorithm makes the result more precise than the general particle filter.