Asymptotically stable phase synchronization revealed by autoregressive circle maps

Abstract

A new type of nonlinear time series analysis is introduced, based on phases, which are defined as polar angles in spaces spanned by a finite number of delayed coordinates. A canonical choice of the polar axis and a related implicit estimation scheme for the potentially underlying auto-regressive circle map (next phase map) guarantee the invertibility of reconstructed phase space trajectories to the original coordinates. The resulting Fourier approximated, Invertibility enforcing Phase Space map (FIPS map) is well suited to detect conditional asymptotic stability of coupled phases. This rather general synchronization criterion unites two existing generalisations of the old concept and can successfully be applied e.g. to phases obtained from ECG and airflow recordings characterizing cardio-respiratory interaction.