An on-line method for segmentation and identification of non-stationary time series

We present a method for the analysis of non-stationary time series from dynamical systems that switch between multiple operating modes. In contrast to other approaches, our method processes the data incrementally and without any training of internal parameters. It straightaway performs an unsupervised segmentation and classification of the data on-the-fly. In many cases it even allows to process incoming data in real-time. The main idea of the approach is to track and segment changes of the probability density of the data in a sliding window on the incoming data stream. An application to a switching dynamical system demonstrates the potential usefulness of the algorithm in a broad range of applications.