Abstract:
We have recently proposed an algorithmic scheme [22] for the non-subjective computation of correlation dimension from time series data. Here it is extended for the computation of generalized dimensions and the multifractal spectrum and applied to a number of EEG and ECG data sets from normal as well as certain pathological states of the brain and the cardiac system. Comparisons are drawn using a standard low dimensional chaotic system. Our method has the advantage that the analysis is done under identical prescriptions built into the algorithm and hence the comparison of resulting indices becomes non-subjective. This also enables a quantitative characterization of the relative complexity between practical time series such as, those corresponding to the changes in the physiological states of the same system, from the view point of underlying dynamics.