小波变换在时间序列特征提取中的应用

时间序列的数学特征主要包括随机项、周期项以及趋势项.针对变形监测等测绘领域中如何对时间序列进行特征提取并进行分析与预测等重要问题,提出一种基于多尺度分析的小波变换方法.首先选取合适的小波基函数、分解层次等参数,其次将待分析的时间序列分解成低频和高频两部分,最后将分解后的时间序列投射到不同尺度上,从而可以提取所需要的有用信息.研究结果表明,基于多尺度分析的小波变换方法能够有效对时间序列进行特征提取,分析出其中的随机项、周期项、趋势项等信息,可以用于GPS变形监测等工程实际中.

Application of wavelet transform to extracting the time series feature

Mathematical characteristics of time series include random items, periodic items and trend items.How to analyze and forecast time series is very important in surveying and mapping, especially indeformation monitoring areas. A multi-scale analysis based on wavelet transform method is proposed.Firstly, the appropriate wavelet function, decomposition level and other parameters are selected;secondly, the time series are decomposed into two parts, the low frequency and the high frequency; lastly,the decomposition of the time series are projected into different scales, which can extract usefulinformation needed. The results show that the wavelet transform based on multi-scale analysis of timeseries can effectively extract the random items, periodic items, trend items and other information items. Itcan be used in engineering practice such as GPS deformation moniloring areas, which has certainapplication values.