scientific article; zbMATH DE number 1193397

From MaRDI portal
Publication:3841874





Fractal models have found wide acceptance in many fields of science. They provide a powerful tool for the study of systems that demonstrate long-term correlations and \(1/f\)-type spectral behavior. The same properties that make fractals powerful models are those that also complicate analysis, however. The extended correlation structure is difficult to capture with standard techniques (e.g., finite-order autoregressive moving average (ARMA) models), and frequency-domain analysis is complicated by the nonstationary character of many fractal processes. Nevertheless, a model known as fractional Brownian motion (fBm) has provided a mathematical framework for the development of analytical methods appropriate for such processes. Fractional Brownian motion is a nonstationary random process that has infinitely long-run correlations and demonstrates \(1/f\)-type spectral behavior. Fractional Brownian motion also has the interesting property of statistical self-similarity, obeying the relationship: \[ \text{var}[B_H(t_2)- B_H(t_1)]\propto (t_2- t_1)^{2H}, \] where \(B_H(t)\) is the fBm process, and \(H\) is the Hurst exponent, a single parameter that characterizes the scaling relationship. In this sense, fBm is fractal; it possesses no characteristic time-scale, and any selected segment properly rescaled is statistically equivalent to the original process. The fBm model has been successfully applied in many fields of science, including physiology. Fractional Brownian motion models have been proposed for signals such as the electroencephalogram (EEG), heart rate variability (HRV), and regional coronary flow distributions. We begin by reviewing the fBm model and the process obtained by calculating the increments of fBm, discrete fractional Gaussian noise. We then review three estimators for \(H\). These estimators, based upon the maximum likelihood estimation, the power spectral density, and the wavelet transform, exploit the scaling behavior in different ways to estimate \(H\). For the entire collection see [Zbl 0889.00018].




Cited in
(18)








This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3841874)