A novel GA multiple model prediction approach with application to system identification driven by chaotic signals

In this paper, we propose a novel multiple model prediction approach using a genetic algorithm (GA). The motivation relies on the fact that many real-life time series cannot be accurately described by a single dynamic model because these time series are composed of more than one underlying regimes along the time scale. Based on a hidden Markov process, the proposed multiple model (MM) is able to capture the temporal relationship among the underlying regimes. A genetic algorithm is employed to train the multiple model and to obtain an optimal segmentation of the time series. Combined with a nonlinear prediction method, this named GA MM predictor is also proposed to identify systems with input signals composed of multiple chaotic dynamics. Applied to a newly developed time division multiuser chaotic communication system, the GA MM approach provides satisfactory channel equalization performance even when the measurement noise is strong.