一种基于代价参考粒子滤波器组的非线性调频信号估计方法  

Cost-reference particle filter bank for nonlinear FM signal estimation

作　　者：卢锦 陶筱娇 LU Jin;TAO Xiao-jiao(School of Electronic Information and Artificial Intelligence,Shaanxi University of Science&Technology,Xi'an 710021,China)

机构地区：[1]陕西科技大学电子信息与人工智能学院,陕西西安710021

出　　处：《陕西科技大学学报》2021年第5期174-179,共6页Journal of Shaanxi University of Science & Technology

基　　金：国家自然科学基金项目(61801281);陕西省教育厅专项科研计划项目(JK170084)。

摘　　要：为提高背景噪声统计特性未知情况下非线性调频信号瞬时频率的估计精度和速度,本文提出一种具有完全并行结构的代价参考粒子滤波算法:代价参考粒子滤波器组.该方法包括以下几个步骤:首先,将非线性调频信号近似为分段线性调频信号,建立状态空间模型;随后,基于状态空间模型和先验信息将状态空间划分为若干较小的子空间;在不同的状态子空间内分别执行代价参考粒子滤波算法;最后,比较各个代价参考粒子滤波算法的累积代价,将累积代价最小的代价参考粒子滤波算法的估计结果作为最终的瞬时频率估计结果.仿真结果表明,与同类的代价参考粒子滤波算法、前后先代价参考粒子滤波算法相比,本文提出的方法估计精度更高,运行时间更短.该方法可用于雷达目标检测与跟踪等领域.A nonlinear filter bank based on cost-reference particle filter(CRPF)is proposed for improving the performance of estimating the nonlinear frequency modulated(FM)signal with unknown background.The method named CRPF bank consists of several steps.First,the nonlinear FM signal is approximated as piecewise linear FM signal,and the state-space model with unknown statistics is adopted to state the filtering problem.Second,the state space is divided into subspaces by considering the initial information and state-space model.Third,several CRPFs are applied to subspaces in parallel.Finally,the filtering results of the CRPF with minimum cumulate cost is derived as the output of filter bank.Some simulation results indicate that,compared with the state-of-art such as CRPF,forward-backward CRPF,CRPF bank achieves comparable filtering performance and requires much short runtime.CRPF bank can be used in radar target detection and tracking.

关 键 词：非线性调频信号估计 非线性滤波器组 代价参考粒子滤波 状态空间模型 

分 类 号：TP391.[自动化与计算机技术—计算机应用技术]