Statistical classification of chaotic signals

The classification of chaotic signals generated by low-dimensional deterministic models given a dictionary of possible models is considered. The proposed classification methods rely on the concept of “best predictor” of signal. A statistical interpretation of this concept based on the ergodic theory of chaotic systems is presented. A sort of “bootstrapping” estimator of the statistical properties is introduced. The method is validated by numerical simulations. Directions for future research are suggested