混沌通信中的粒子滤波技术研究

【作者】 徐茂格；

【导师】 宋耀良；

【作者基本信息】 南京理工大学 ， 信息与通信工程， 2008， 博士

【摘要】 非线性信号处理一直是信号处理领域的研究热点与难点。除了少数特殊环境,人们大都采用解析近似或者数值计算的方法来解决非线性问题,然而这些方法易陷入局部极值或者面临巨大的计算量。粒子滤波从贝叶斯理论出发采用蒙特卡洛抽样,提供了一种灵活的方式来解决非线性问题,是近年非线性信号处理的重要研究方向。混沌信号是典型的非线性信号,其非线性和长期不可预测等特点导致混沌通信中许多滤波问题变得异常复杂。在混沌通信中,混沌同步与滤波、远距离混沌通信的噪声影响和信道畸变、多用户通信与抗多径传播等问题是目前混沌通信从理论研究转向实际应用中需要解决的关键问题。本论文基于粒子滤波,从理论和算法上对混沌同步、信道估计与均衡、混沌通信信号检测等方面进行了深入研究。混沌通信是建立在传统的通信基础上,传统的通信面临的问题它一样可能会遇到。最后结合现代无线通信系统,讨论了粒子滤波技术应用中的关键技术——降维,主要研究内容和创新如下:1)从自适应滤波的角度研究了混沌同步,阐述了扩展卡尔曼滤波(EKF)在混沌同步中出现退化现象的原因,提出了基于粒子滤波的混沌同步方法。从Cram’er-Raolower bound(CRLB)下界出发提出了一种自适应扰动噪声的方差确定方法,解决了粒子滤波同步方法的难点:扰动噪声的选择问题。2)对于混沌多址通信,因多种混沌信号混合,接收端存在同时分离与同步问题,对此提出了基于粒子滤波的在线盲分离方法。为了降低分离后信号的残留噪声,结合混沌信号降噪思想给出了一种新的延迟估计方法。与传统的延迟加权的估计方法相比,该延迟估计方法能极大的减少延迟时间,提高了计算效率。3)对于平坦衰落信道Jakes模型,基于贝叶斯预测技术,提出了一种新的信道建模方法。主要思想是对AR模型的参数引入随机游动(random walk)的变化。结合该信道动态特性模型,基于粒子滤波,给出了一种稳健的信道估计方法。该方法能减轻欠估计的归一化多普勒频率对信道跟踪的影响。4)针对结合加密函数的混沌掩盖通信体制,提出了联合混沌同步与信号检测的方案。该方案不需要传输额外的同步信号,仅靠含有未知信息序列的接收信号就能引导接收端达到自同步。进一步为了降低算法的复杂性,设计了一种新的重要性函数,该重要性函数充分地利用了传输符号的离散性。5)针对将信息序列掩盖在混沌调频信号中的混沌通信体制,提出了基于粒子滤波的频率跟踪方法,该技术不仅能跟踪简单变化的频率而且能有效地跟踪混沌调频信号的频率。在此基础上推导了混沌调频信号频率跟踪的后验克拉美一罗(PCRB)下界,仿真结果表明粒子滤波有较好的跟踪性能和稳定性。对于混沌调频信号,所提出的方法的频率跟踪均方误差与PCRB在同一个数量级。6)讨论了粒子滤波实用化过程中的关键技术——一般情况下的降维。以MIMO频率选择性衰落情况下联合信道估计与信号检测为背景,提出了一种时延域粒子滤波。主要思想是利用一组粒子滤波在时延域分别估计多径延迟分量,由此降低了单个粒子滤波的采样空间维数,为粒子滤波在高维信号中的估计提供了一种思路。更多还原

【Abstract】 The nonlinear signal processing is the difficult and hot topic in signal processing. Only a few narrow classes of models have exact solutions, and a number of approximate filters have been devised for more generalized cases. However, these traditional filters may be easy to get in local minimum and face the huge computation cost. Combining Bayesian theory with Monte Carlo sampling, particle filtering provides a flexible way to solving nonlinear problems. Chaos signal is the typical nonlinear signal. The nonlinearity and long time unpredictable property of chaotic signal cause the filtering problems of chaos communication difficult to conduct. Chaos synchronization, channel equalization, multi-user signal detection are the key problems in chaotic communications for practical implementation. In this dissertation, based on particle filtering, chaotic signal analysis, channel estimation and chaotic communication signal detection are studied in detail. Extended to regular wireless communication, the important technique for particle filtering-the sampling dimensionality decreasing is also discussed. The main work of my research is as follows:1) We study the chaotic synchronization based on extended Kalman filter (EKF) and particle filtering. We analyze degenerate phenomena of EKF. And a robust chaotic synchronization method based on particle filtering is proposed. Utilizing the Cramer-Rao low bound, an adaptive variance choice strategy for roughing noise is developed.2) In multi-user environments, a online blind separation algorithm based on particle filtering is proposed. Further more, a novel delay estimation method is also suggested, which can effectively reduce the residual noise in the recovered signal compared to the traditional delay-weight method.3)For the flat fading channel environments, based on Bayesian forecasting technology, a time-varying wireless channel model is proposed. Utilizing particle filtering and the channel model, a robust wireless channel tracking scheme is developed. Compared with the traditional tracking scheme, this scheme doesn’t need to know exactly the normalized doppler frequency, and can greatly decrease the modeling error.4) For the chaos masking communication scheme using encryption function, a novel Bayesian receiver is provided. Although there is not additional synchronization signal, the proposed technique can easily synchronize the chaotic system in transmitter utilizing the received signal which has unknown information. Further more, in order to decrease the complexity and improve the performance of the proposed receiver, a suboptimal importance function is suggested, which combines the prior distribution of chaotic state and the posterior distribution of information symbols.5) For the chaos communication scheme which masks the information signal in the chaotic frequency modulation signal, particle filtering for frequency tracking is introduced, and also its feasibility is analyzed. The posterior Cramer-Rao Bounds for the frequency tracking of the chaotic frequency modulation signal is also derived. The simulation demonstrates the superiorities of particle filtering.6) The key technique of particle filtering- the sampling dimensionality decreasing for the general environment is studied. In the background of MIMO frequency selective channel estimation, a time delay domain particle filtering is suggested. The main idea is to change the time domain channel estimation into time delay domain processing, and a bank of particle filters is utilized, thus the sampling dimensionality is much small for each particle filters.更多还原