Distinguishing Chaotic Time Series from Noise: A Random Matrix Approach
Bin Ye,Jianxing Chen,Chen Ju,Huijun Li,Xuesong Wang
DOI: https://doi.org/10.1016/j.cnsns.2016.08.018
IF: 4.186
2017-01-01
Communications in Nonlinear Science and Numerical Simulation
Abstract:Deterministically chaotic systems can often give rise to random and unpredictable behaviors which make the time series obtained from them to be almost indistinguishable from noise. Motivated by the fact that data points in a chaotic time series will have intrinsic correlations between them, we propose a random matrix theory (RMT) approach to identify the deterministic or stochastic dynamics of the system. We show that the spectral distributions of the correlation matrices, constructed from the chaotic time series, deviate significantly from the predictions of random matrix ensembles. On the contrary, the eigenvalue statistics for a noisy signal follow closely those of random matrix ensembles. Numerical results also indicate that the approach is to some extent robust to additive observational noise which pollutes the data in many practical situations. Our approach is efficient in recognizing the continuous chaotic dynamics underlying the evolution of the time series. (C) 2016 Elsevier B.V. All rights reserved.