Bivariate global frequency analysis versus chaos theory. A comparison for sleep EEG data

Various quantitative descriptors for EEG data will be compared taking sleep as an example. In this contribution, Hjorth's mobility and complexity measures will be used to classify sleep stages. The results will be compared with those of a dimensionality analysis. Several authors have shown that the correlation exponent can describe the complexity of sleep EEG data and is able--with the exception of REM sleep--to distinguish significantly between sleep stages. The discriminative power of a bivariate global frequency analysis appears to be superior to that of the correlation exponent. Furthermore a very high statistical correlation between the estimator of fractal dimension and Hjorth's mobility was obtained.