机构地区: 西安电子科技大学电子工程学院电子工程系
出 处: 《信号处理》 1997年第2期97-104,共8页
摘 要: 本文我们主要讨论拓扑维为一维的时间序列的分形分析和参数提取,着重论述分维数的估计,介绍了几种常用方法,并提出了一种基于有限长度的方差估计法,用实验分析了这些计算方法的准确性、平稳性、计算复杂度、受噪声和信号长度的影响等方面的性能,得到一些有意义的结论,实验结果表明,我们提出的方法在某些方面具有优点。 In this paper, we deal with the problem of fractal-based analysis of and features extraction from real- valued topological 'one-dimensional' time-series data, with emphasis on the estimation of fractal dimension. We introduce several commnon methods and propose a new algorithm to measure fractal dimentsion, Experiments are used to analyse their performaces such as estimation accuracy、 stability、 computaion complexity and the effects of added noise and length of series on all these methods, some useful conclusions are resulted, which show the proposed method is efficient and somewhat superior.