基于DSP的改进粒子滤波算法探究.pdfVIP

基于DSP的改进粒子滤波算法研究 凤俊翔，张健，薄超 (江苏科技大学电子信息学院，江苏镇江212003) 摘 要：介绍了粒子滤波基本原理，针对粒子滤波计算量大和难以用硬件实现等缺点对粒子滤波算法进行了改进，使其 平均计算周期缩短为原来的90％，应用DSP实现了粒子滤波算法。改进粒子滤波算法主要优化了原粒子滤波算法中权值 计算、重采样和输出步骤，使其计算速度和滤波精度有所提高。这种改进粒子滤波算法在DSP系统中进行仿真，结果证明它 具有速度快，精度高的优点。 关键词：粒子滤波；硬件实现；权值计算；数字信号处理器 中图分类号：TN911-34 文献标识码：A 文章编号：1004—373X(2010)18-0009—04 ParticleFilter BasedonDSP Improved AIgorithm FENG Jian，130Chao Jun-xiang，ZHANG (Sch001of and ElectronicsIMormation．JiangsuUniversityofScienceand 212003，China) Technology，Zhenjiang filterisbasedontheMonteCarloandrecursive estimationwhichhas in Abstract：Particle Bayesian specialadvantages withthenonlinearandnon—Gaussian theenormous andlow restrictits dealing problems．However，bothcomputationsspeed inreal—time basic of filteris filter first，the introduced．Then，the implementationsystem．At theoryparticle particlealgo— rithmis for the ofenormous andhard ofhardware，whichre— improvedsolvingdisadvantages computationsimplementation ducedthe to filter isfulfilled with fil— average 90％．At algorithm through cycle last·particle DSP．Comparedoriginalparticle ter the of of and the algorithm，theimprovedalgorithmoptimizesstepscomputationweights，resampleoutput，whichimprove calculatorandfilter filter testifiesits offast and accu— improved advantages h