基于matlab的功率谱分析方法研究
摘 要
数字信号处理(DSP)重要的应用领域之一,是建立在周期信号和随机信号基础上的功率谱估计。在实际应用中往往不能获得具体信号的表达式,需要根据有限的数据样本来获得较好的谱估计效果,因而谱估计被广泛的应用于各种信号处理中[1]。
本论文研究了功率谱估计的几种常用的方法,包括经典谱估计和现代谱估计的各种方法,且对每种方法的估计质量做了数学推导,并给出仿真程序及仿真图。经典法主要包括周期图法、自相关法,但这两种方法都存在缺陷,即认为观测数据之外的数据都为零,所以对经典法中的周期图法进行了加窗、平均等修正,因此提出了周期图法的改进方法;现代谱估计的方法分类比较多,AR模型法,MA模型法和ARMA模型法是现代功率谱估计中最主要的参数模型,本论文着重讨论了AR模型参数法[2]。同时论文将通过对经典谱估计和现代谱估计的实现方法及仿真图的比较,得出经典功率谱估计方法的方差性较差,分辨率较低,而现代谱估计的目标正是在于努力改善谱估计的分辨率,因此能得到较好的谱估计效果,为此应用更为广泛[3]。
关键字:数字信号处理,功率谱估计,周期图法,自相关法,AR模型法
ABSTRACT
Digital signal processing (DSP) important application of one of the field. Actually, we can’t get the expression of a specific signal, so we need to estimate the power spectral of a signal according to some sample data sequences.So spectrum estimation which is widely used in various signal processing.
In this thesis, some common methods of Power Spectral Estimation, such as classical spectral estimation and modern spectral estimation, are studied. The quality of each estimation method is derived, simulation program and simulation figure is given. Classical methods of Power Spectral Estimation mainly include the Periodogram and the BT method. But both of them have a common drawback: the data sequences, beyond the area of the observed sequences, are all presumed to zero. So the Windows and the average method are introduced to improve the quality of the Periodogram. Therefore the improvement of The Periodogram estimation method is proposed. The classification of modern spectral estimation methods are more , AR,MA, and ARMA is the most important parameters of modern spectral estimation. This thesis will focus on discussion of AR model parameters method. At the same time , It can be seen from the comparison and realization of classical spectral estimation and modern spectral estimation, classical power spectrum estimation variance is poor, low resolution .The goal of modern spectral estimation is working to improve the resolution of spectral estimation, better results