【作者】 杨强；

【导师】 山拜·达拉拜；

【作者基本信息】 新疆大学 ， 通信与信息系统， 2011， 硕士

【摘要】 信号处理是信息传输科学中近十几年发展最快的学科之一,传统的信号处理有三个基本的假设:线性,高斯性和平稳性;而现代信号处理方法是以非线性,非高斯性和非平稳性作为分析与处理的对象,尤其以带有非高斯噪声的信号处理引人注目。本文研究是建立在双模噪声基础上的通信和信号处理理论进行了完善和补充。从而研究了多模噪声,实质是多种噪声迭加产生的混合噪声,属于非高斯噪声。非高斯噪声的研究方法一般是高阶统计量法,本文采用了现代信号处理的方法。主要工作如下:(1)综述了通信中信号处理发展现状,介绍了噪声处理技术历史和发展的现状,并且说明了多模噪声研究的理论意义。(2)系统地研究和论述了各种非高斯噪声理论,然后对不同类型的噪声信号进行了分析,实现有用信号和噪声分离。系统地阐述了统计信号检测基本理论和统计信号的判决准则。(3)研究和分析了双模噪声模型,引出了多模混合噪声的四种模型,深入分析了多模噪声的统计特性中绝对值均值和平均功率,为以后的研究提供了基础。(4)研究了自适应算法和LMS Newton算法,提出了改进算法,使得噪声和信号很好的分离。(5)研究了基于贝叶斯理论的信号估计,在实际问题中,利用带有噪声的观测量对系统状态进行滤波与估计,常常采用状态空间法对系统建模。本文提出了一种结合多层感知器(MLP)和粒子滤波融合算法。更多还原

【Abstract】 In the past decade, the processing of signal has been one of the fastest developing subjects. The traditional signal processing is based on three suppose: linearity, Gaussianity and stationarity. While the modern signal processing non-linearity, non-Gaussianity and non-stationarity. Especially, the signal with the non-Gaussian noise has caught most people’s eyes.This paper gives some perfection and complement to the theory of communication & signal processing based on bimodal hybrid noise. Then multi-modal noise has been propose, and is a hybrid noise consisting of many kinds of additive noises. It belongs to non-Gaussian noise. Generally,the method used to study non-Gaussian noise is using high-order statistics(HOS). The paper uses the method of modern signal processing. Main content of this paper is as follows:(1)It reviews present conditions of development of communication systems, and introduces the history and current situation of the processing technology of noise, and provides the perspective of the theory of multi-modal noise.(2)The thesis systematically studied and discussed a variety of non-Gaussian noise theory. Then illustrate all kinds of noise and put forward exact algorithm to separate the noise and signal. We also systematically described the basic theory of statistical signal detection and the adjudging rules of statistical signal.(3) Research and analysis of the bimodal noise, leads to the four models of the multi-modal hybrid noise. Then in-depth analyze the statistical properties of multi-modal noise in the absolute mean and average power. So it is the foundation for future research.(4)Research of the adaptive algorithm and LMS Newton algorithm, and the improved algorithm has been proposed, which is making the noise and signal well separated.(5)It studies estimation based on Bayesian theory, in many problems, used observed value with noise to filter and estimate on systems status, people often adopt state space method to simulated system. The paper provides a kind of particle filter with MLP fusion to realize to signal estimation.更多还原