阈值去噪下的改进粒子滤波算法

Improved Particle Filter Algorithm Based on Threshold De-Noising

针对粒子滤波在非线性系统上具有优越性,但粒子在传播过程中必然受到噪声影响的问题,提出了在阈值去噪下的改进粒子滤波算法.将小波阈值去噪的思想引入到粒子滤波中,即信号先经过小波包分解,再利用适当的阈值保留分解系数较大者并将系数较小者置为0,这样每个粒子结合其历史信息可降低噪声水平,进而改进滤波的状态估计值.蒙特卡罗仿真实验表明,加入阈值去噪的粒子滤波法可以有效降低滤波的均方根误差,提高滤波精度.在所采用的线性及非线性系统中,均方根误差均值分别降低了14%和12%.

Concerning the advantage of particle filter in non-linear systems, and the problem of noise influence during the dissemination process, an improved particle filter algorithm with threshold de-noising is proposed. The idea of wavelet threshold de-noising is introduced into the particle filter, that is, after applying the wavelet packet decomposition on the signal, the coefficients that are greater than a certain threshold are reserved while other coefficients are set to 0. Thus, with the assistance of particle history information, the noise is reduced and the estimated filter state value is more accurate. Monte Carlo simulation results show that the particle filter with threshold de-noising can effectively reduce the filter root mean square error （RMSE） and improve filter accuracy, and that RMSEs in the corresponding linear and non-linear systems are reduced by 14% and 12% respectively.