C. Diks, ESTIMATING INVARIANTS OF NOISY ATTRACTORS, Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics, 53(5), 1996, pp. 4263-4266
We propose a method for estimating the correlation dimension and corre
lation entropy of a time series. It is based on a generalization of th
e correlation integral that is specifically useful when the time serie
s is corrupted with Gaussian measurement noise. From computational exp
eriments we conclude that reasonable estimates for the noise level, co
rrelation dimension, and correlation entropy can be found for time ser
ies with up to 20% noise. The method appears to be fairly robust with
respect to the noise distribution.