Multifractal mass exponent spectrum of complex physiological time series

Abstract

Physiological signal belongs to the kind of nonstationary and time-variant ones. Thus, the nonlinear analysis methods may be better to disclose its characteristics and mechanisms. There have been plenty of evidences that physiological signal generated by complex self-regulated system may have a fractal structure. In this work, we introduce a new measure to characterize multifractality, the mass exponent spectrum curvature, which can disclose the complexity of fractal structure from total bending degree of the spectrum. This parameter represents the nonlinear superpositions of the discrepancies of fractal dimension from all adjacent points in the curve and therefore solves the problem of original parameters for not fully reflecting the information of entire subsets in the fractal structure. The evaluations of deterministic fractal system Cantor measure validate that it is completely effective in exploring the complexity of chaotic series, and is also not affected by nonstability of the signal as well as disturbances of the noises. We then apply it to the analysis of human heart rate variability (HRV) signals and sleep electroencephalogram (EEG) signals. The experimental results show that this method can be better to discriminate cohorts under different physiological and pathological conditions. Compared with the indicator of singularity spectrum width, there are some improvements both on the computing efficiency and accuracy. Such conclusion may provide some valuable information for clinical diagnoses.