Symbolic Dynamic Analysis of Physiological Time Series

An efficient nonlinear analysis method is proposed to characterize the dynamics of physiological time series. This method consists of analyzing the symbolic dynamics of the reconstructed phase space of a time series. Since a physiological time series is usually nonstationary, to compensate for the time varying local mean and extract the wave characteristics of the time series, all the vectors in the phase space are normalized. The maximum topological entropy (WE) criterion is then introduced to find a partition of the phase space. Assessment of this partitioning technique is made using the logistic map and postural sway signals. We used two measures from symbolic dynamics to characterize the dynamics of the original time series. The calculated results for the postural sway signals show that this method enables detecting the dissimilarity of physiological time series in different physiological states.