We present an approach for real-time change detection in the transient phases of complex dynamical systems based on tracking the local phase and amplitude synchronization among the components of a univariate time series signal derived via Intrinsic Time scale Decomposition (ITD)--a nonlinear, non-parametric analysis method. We investigate the properties of ITD components and show that the expected level of phase synchronization at a given change point may be enhanced by more than 4 folds when we employ multiple intrinsic components. Next, we introduce a concept of maximal mutual agreement to identify the set of ITD components that are most likely to capture the information about dynamical changes of interest, and define an InSync statistic to capture this local information. Extensive numerical as well as real-world case studies involving benchmark neurophysiological processes and industrial machine sensor data suggest that the present method can detect sharp change points and second/higher order moment shifts with average sensitivity of ~91% as compared to approximately 30% for other contemporary methods tested on the case studies presented.

plus-circle Add Review

comment

Reviews