缺少观测数据下信号谱分析

Signal Spectral Analysis of Missing Observation

提出了一种处理时间序列中出现数据丢失时的信号谱估计的方法。这时观测所得到的,不再是连续等间隔的时间序列,而是多个数据段,所要进行的即是对这些分段数据的自回归模型的估计。该方法基于标准Burg谱估计算法提出,算法可以建立一个同时适用于各个分段数据的统一的信号模型。在仿真部分的结果显示,与直接使用均值方法进行谱估计相比较,分段Burg算法偏差更小,谱估计更精确。

A method is proposed to find the power spectrum of signals with missing observations. Here we often get several segmented data,but not a consecutive time series, The estimation of an AR model for this type of data is then discussed, The method is based on the standard Burg spectral estimation and gives some modification to find a signal model fitted for all segments simultaneously, As a result,comparing to the averaging approach,the model estimated by the Burg for segments is less biased and more precise,