Study Of Sound Tracking Based On Intelligent Particle Filter

Traditional particle filter performs better than Kalman filter in nonlinear systems,but particle degradation also affects the performance of particle filter.Intelligent particle filter can make low-weight particles evolve into high-weight particles,which can improve the particle diversity to some extent.On the basis of intelligent particle filter,the idea of adaptive processing and particle swarm optimization is added to further optimize genetic resampling and improve the performance of intelligent particle filter in sound source tracking.Adaptive intelligent particle filter,starting from the particle mutation probability,its mutation strategy is still used in intelligent particle filter.Based on the idea of particle swarm optimization,the improved intelligent particle filter optimizes the mutation operator from the angle of the particle mutation direction,and the particle adjusts the mutation direction according to the posterior probability distribution.The particle filter,extended Kalman filter,regular particle filter,intelligent particle filter,adaptive intelligent particle filter and improved intelligent particle filter are simulated through standard particle filter validation model,free-fall motion model and free-fall under nonGaussian noise system.The root-mean-square-error,average error,effective particle number,average run time and tracks are counted.The algorithm is analyzed from several perspectives.The simulation results show that the performance of the improved intelligent particle filter in many sides is due to the intelligent particle filter,especially the intelligent particle filter based on the idea of particle swarm optimization.In many models,the performance of the improved intelligent particle filter can be improved by 10%-60%.In the highly linear free fall simulation,the statistical results of the Intelligent Particle Filter Based on the idea of particle swarm optimization are still better than the extended Kalman filter algorithm,and the speed estimation accuracy is greatly improved.However,the complexity of the algorithm also makes the algorithm consume more computing resources,the average running time is slightly higher than that of intelligent particle filter and adaptive particle filter,but much lower than that of regular particle filter,which guarantees a certain timeliness and improves the performance.The target motion tracking simulation based on TDOA is designed to simulate the various trajectories of the target with system noise,to simulate the effects of noise and reverberation in the source signal with TDOA noise,and to verify the robustness of the algorithm with different size of TDOA noise.The results show that no matter how many TDOA values are taken,the improved intelligent particle filter has higher performance advantages,smaller error fluctuation and better convergence.At the same time,it is also better than the intelligent particle filter in robustness of the algorithm.The performance improvement of the adaptive particle filter is only slightly lower than the improved intelligent particle filter.The intelligent particle filter has good tracking results,but its error fluctuation is large.In the distributed array target tracking experiment,the trajectory of the target is designed with multiple sets of motion equations to verify the engineering feasibility of the improved strategy.The results show that the improved intelligent particle filter based on the idea of particle swarm optimization has a much better tracking performance than other methods,and the adaptive particle filter can keep a certain tracking accuracy.