Sleep-stage Characterization by Nonlinear EEG Analysis using Wavelet-based Multifractal Formalism.

The wavelet-based multifractal formalism was applied on sleep EEG analysis and sleep-stage characterization. The subjects used in this study were randomly selected from the MIT-BIH Polysomnographic Database. The multifractal singularity spectra of sleep EEG signals were estimated, and h<inf>0</inf>, the Hölder exponent that denotes the main singular property of the signal, was extracted from the multifractal singularity spectrum and used as sleep-stage characteristic parameter. The shift of multifractal singularity spectra of different sleep stage was observed. The mean h<inf>0</inf>exponents increased from awake to sleep stage 1, 2, 3 and 4, but decreased during rapid eye movement (REM) sleep. Our study suggests that the h<inf>0</inf>exponent could be used as an important sleep stage characteristic parameter.