论文总字数:18940字
摘 要
功率谱估计是数字信号处理中一种常见的研究方法,可以获得随机信号功率随频率的变化情况。而功率谱估计分为两类方法,一类是经典功率谱估计,另一类是现代功率谱估计。其中经典功率谱估计以基于FFT变换的自相关法为例,是一种非参数化的估计方法,而现代功率谱估计中则使用了不同的模型对参数进行估计,是一种参数化的估计方法。
本文即以自相关法与ESPRIT算法为例,对相同的噪声环境分别进行了两种功率谱估计的研究,对比了非参数化与参数化估计的特点,并使用MATLAB进行了以下仿真实验:
- 改变仿真信号中若干单频信号的频率、数目、幅度,观察发现ESPRIT算法具有稳定的功率谱估计性能;
(2)改变仿真信号中的噪声强度,分析算法能够准确进行谱估计的信噪比范围,发现ESPRIT算法在高信噪比下表现良好,在低信噪比时,ESPRIT算法受到噪声的影响较大,与真实值还有一定的差距;
(3)不断减小某两个点频的距离,查看算法的功率谱分辨率,即能够将间隔多大的两个点频在谱域分开,研究发现ESPRIT算法分辨率明显优于FFT算法;
(4)不断缩短仿真信号的长度,并比较在不同数据长度下,功率谱估计的仿真结果。结果表明,当采样数据变少时,对参数的估计偏差较大。
关键词:功率谱估计;ESPRIT;FFT;参数估计;
Random signal power spectrum estimation based on ESPRIT algorithm
Jingyi Xu, 04015632
Supervised by Zhongjin Jiang
ABSTRACT
Power spectrum estimation is a common research method in digital signal processing, and the random signal power can be obtained as a function of frequency. Power spectrum estimation is divided into two types of methods, one is classical power spectrum estimation, and the other is modern power spectrum estimation. The classical power spectrum estimation is based on the autocorrelation method based on FFT transform, which is a nonparametric estimation method. In modern power spectrum estimation, different models are used to estimate the parameters, which is a parameterized estimation. method.
In this paper, the autocorrelation method and the ESPRIT algorithm are taken as examples. Two power spectrum estimations are studied for the same noise environment. The characteristics of nonparametric and parametric estimation are compared. The following simulation experiments are carried out using MATLAB:
(1) Change the frequency, number and amplitude of several single-frequency signals in the simulated signal, and observe that the ESPRIT algorithm has stable power spectrum estimation performance;
(2) Changing the noise intensity in the simulated signal, the analysis algorithm can accurately calculate the signal-to-noise ratio range of the spectrum estimation, and find that the ESPRIT algorithm performs well at high SNR. At low SNR, the ESPRIT algorithm is affected by noise. Large, there is still a certain gap with the true value;
(3) Constantly reduce the distance between two points and look at the power spectrum resolution of the algorithm, that is, the two point frequencies of the interval can be separated in the spectral domain. It is found that the resolution of the ESPRIT algorithm is significantly better than the FFT algorithm;
(4) Constantly shorten the length of the simulation signal and compare the simulation results of power spectrum estimation under different data lengths. The results show that when the sampling data becomes less, the estimation bias of the parameters is larger.
Keywords: Power spectrum estimation; ESPRIT algorithm; FFT; Parameter estimation;
目 录
摘要
ABSTRACT
第1章 绪论
1.1引言
1.2研究现状
1.3论文结构
第2章 经典谱估计理论
2.1信号模型
2.2基于FFT的两种分析方法
2.2.1周期图法
2.2.2自相关法
3本章小结
第3章 ESPRIT算法
3.1阵列信号处理模型
3.2 ESPRIT数据模型
3.2.1窄带信号
3.2.2阵列流型
3.2.3旋转不变性
3.3 基于ESPRIT的功率谱估计
3.3.1欧拉公式变形
3.3.2旋转矩阵的变形
3.3.3 TLS-ESPRIT总体最小二乘理论
3.3 本章小结
第4章 参数化ESPRIT与非参数化FFT的仿真比较
4.1引言
4.2功率谱估计性能
4.3频谱分辨率
4.4不同信噪比下的误差
4.5信号长度
4.6本章小结
第5章 总结与展望
5.1总结
5.2展望
参考文献
致 谢
第1章 绪论
1.1引言
随机信号是一种在生活中十分常见的信号。这些信号在通信的传输过程中,信号往往受到一定的外界干扰和噪声,这些噪声往往都是随机的,具有不确定性,这就使接收的信号也具有了不确定性。对于不确定的信号,使用频谱方法的分析是不合适,因为频率在不断地进行改变,必须使用统计的方法来分析信号源的特性,而对其功率谱进行研究就是其中一种方法。
功率谱如今已经被广泛地应用于各个领域,如通信、电力、地质勘探、医学、天文等,从功率谱的分析中,可以得知很多重要的频率和功率特性。
剩余内容已隐藏,请支付后下载全文,论文总字数:18940字
该课题毕业论文、开题报告、外文翻译、程序设计、图纸设计等资料可联系客服协助查找;