摘要
为克服传统的周期图谱估计方法存在的分辨率低及非一致估计等缺点而迅速发展的各种近代谱估计法只能适用于ＡＲＭＡ（Ａｕｔｏｒｅｇｒｅｓｓｉｖｅ－ｍｏｖｉｎｇａｖｅｒａｇｅ）信号的谱估计。为解决存在于各种物理过程中的非ＡＲＭＡ随机过程（包括１／ｆ过程）的谱估计，本文提出了一种对非ＡＲＭＡ随机信号的非递归数字滤波算法，重点研究了算法主要参数（滤波器长度、数据长度）的选择原则及对谱估计准确度和精度的影响。理论分析及实际测量证明，此方法具有分辨率高、谱估计精确较好及一致估计等优点，特别适用于对非ＡＲＭＡ信号的某些特征频率点的谱估计。
he modern spectral estimation used to overcome the lower resolution and inconsistency of periodogram approach can only be utilized in the spectral estimation for ARMA stochastic signal. In this paper, a flew non-recursive digital filter algorithm used in the spectral estimation for a non-ARMA stochastic process (including 1/f process) is proposed. The selection principle for the major parameters of algorithm (the filter length and data length) and its effect on the precision and accuracy of spectral estimation are given. The theoretical analysis and measurement results show that this method has higher resolution and precision, particularly, it may be used to estimate the spectral value at some characteristic frequency of a non-ARMA stochastic process.
出处
《计量学报》
CSCD
1994年第2期86-91,共6页
Acta Metrologica Sinica
基金
国家自然科学基金
关键词
随机信号
谱估计
数字滤波
算法
Stochastic signal
Spectral estimation
Digital filter algorithm
