基于matlab的ar模型谱估计研究与实现-毕业论文设计.doc

摘 要 信号的频谱分析是研究信号特性的重要手段之一，对于确定性信号，可以用Fourier变换来考察其频谱性质，而对于广义平稳随机信号，由于它一般既不是周期的，又不满足平方可积，严格来说不能进行Fourier变换，通常是求其功率谱来进行频谱分析。功率谱估计在近30年中获得了飞速发展。涉及到信号与系统、随机信号分析、概率统计、随机过程、矩阵代数等一系列学科，广泛应用于雷达、声纳、通信、地质、勘探、天文、生物医学工程等众多领域。 实际中，数字信号的功率谱只能用所得的有限次记录的有限长数据来予以估计，这就产生了功率谱估计这一研究领域。功率谱的估计大致可分为经典功率谱估计和现代功率谱估计经典谱估计的两个主要方法为周期图法和自相关法。针对经典谱估计的分辨率低和方差性能不好等问题提出了现代谱估计现代谱估计大致可以分为参数模型谱估计和非参数模型谱估计基于参数建摸的功率谱估计是现代功率谱估计的重要内容，其目的就是为了改善功率谱估计的频率分辨率，它主要包括AR模型、MA模型、ARMA模型，其中基于AR模型的功率谱估计是现代功率谱估计中最常用的一种方法。 理论分析及MATLAB仿真结果表明：经典谱估计方法得到的功率谱出现了许多虚假的谱峰，频率分辨率很低，而现代谱估计方法得到的功率谱较为真实，没有明显的频率偏移和假峰，并且具有较高的频率分辨率，尤其是频率带宽性能得到了明显的改善。 关键词：功率谱估计；AR模型；MATLAB；Levinson-Durbin算法；Burg算法 ABSTRACT Signal spectral analysis is one of the most important means to examine the characteristics of signal. Fourier transform can be used to study the of the spectrum of the certainty signal. For general stochastic signal, it is neither a cycle in general, nor in line with the square integration .Strictly speaking, general stochastic signal cannot be transformed by Fourier transform. So the power spectrum is generally used for signal spectral analysis. In the last 30 years Power spectral estimation was rapidly developed. It related to a range of disciplines such as Signals and systems, stochastic signal analysis, probability and statistics, stochastic processes and Matrix algebra. And it is widely used in radar, sonar, communications, geology, exploration, astronomy, biomedical engineering and many other fields. Actually, the power spectrum of digital signal can only be estimated by finite length data derived from the limited records, which produced the study area of power spectrum estimation. Power spectral estimation can be broadly divided into classical power spectral estimation and modern power spectral estimation. Two main methods of Classical power spectral estimation are period gram method and auto-correlation method. For the issues such as low resolution and poor variance performance in Classical spectral estimation, modern spe