Clustering of time-evolving scaling dynamics in a complex signal

Complex time series are widespread in physics and physiology. Multifractal analysis provides a tool to study the scaling dynamics of such time series. However, the temporal evolution of scaling dynamics has been ignored by traditional tools such as the multifractal spectrum. We present scaling maps that add the time dimension to the study of scaling dynamics. This is particularly important in cases in which the dynamics of the underlying processes change in time or in applications that necessitate real-time detection of scaling dynamics. In addition, we present a methodology for automatic clustering of existing scaling regimes in a signal. We demonstrate the methodology on time-evolving correlated and uncorrelated noise and the output of a physiological control system (i.e., cardiac interbeat intervals) in healthy and pathological states.