Spectral Detrended Fluctuation Analysis and Its Application to Heart Rate Variability Assessment

摘要

Detrend fluctuation analysis (DFA) has become a choice method for effectiveanalysis of a broad variety of nonstationary signals. We show in the presentarticle that, provided the nonstationary fluctuations occur at a large enoughtime scale, an alternative approach can be obtained by using the Fourier seriesof the signal. More specifically, signal reconstructions considering Fourierseries with increasing number of higher spectral components are subtracted fromthe signal, while the dispersion of such a difference is calculated. The slopeof the loglog representation of the dispersions in terms of the time scale(reciprocal of the frequency) is calculated and used for the characterizationof the signal. The detrend action in this methodology is performed by the earlyincorporation of the low frequency spectral components in the signalrepresentation. The application of the spectral DFA to the analysis of heartrate variability data has yielded results which are similar to those obtainedby traditional DFA. Because of the direct relationship with the spectralcontent of the analyzed signal, the spectral DFA may be used as a complementaryresource for characterization and analysis of some types of nonstationarysignals.