时序资料的时间窗平滑法及应用

观测资料的平滑处理，是数据分析中一类重要的常用方法，平滑处理的目的，在于既保留资料中的主要有用信号，又消除或削弱随机误差，本文对密集采样时序资料，提出了时间窗平滑法，并以光谱和光电测量资料为例，通过分析ＰＤＭ技术中拟合曲线的计算，说明了它的性质与应用，与高斯权函数平滑法相比，此方法计算简便，且有更明确的物理意义，与观测值多点合并法相比，有更好的采样一致性和避免损失实际的分辨率，当用于等间距时序资料

TIME-WINDOW SMOOTHING METHOD AND ITS APPLICATIONS

In data analysis,smoothing reduction of the observation data is a commonly used method. This processing can remove or largely reduce random noisy and,at the same time, keep the most useful signals in the data. For a time series unequally spaced data, a new method, Time-Window smoothing, is suggested in this paper.which has the form where Pi is a measurement weight of the datum f(ti),W, is the width of the Time-Window.With the examples in reduction of spectral observations and in analysis of photoelectrical measurements, and also with the dissection of how to get a fitting curve in PDM technique,the characters and applications of the Time-Window smoothing are illustrated. Compared to Gaussian weighting function smoothing, it may have almost the same effect,but needs only much less computation time and shows clearer and more definite physical meaning.Compared to multipoint-binning method,it may have better uniformity of sampling,and avoid lossing more real resolution.In the case of equally spaced data, this method degenerates to the usual K-point sliding averaging.