Search

This paper provides an insight into dimension analysis from time series. In particular, we propose a procedure based on the pointwise dimension in order to extract, for each embedding dimension, the subset of points in the phase space (and the corresponding ones in the time series) which give rise to the scaling behaviour. We may consider the output time series as the result of a filtering process, based on correlations of points in the phase-space domain. Furthermore, the procedure gives the statistics of points which determine the scaling behaviour. [Copyright &y& Elsevier]

We present and compare two automatic methods for dimension estimation from time series. Both methods, based on conceptually different approaches, work on the derivative of the bi-logarithmic plot of the correlation integral versus the correlation length (log–log plot). The first method searches for the most probable dimension values (MPDV) and associates to each of them a possible scaling region. The second one searches for the most flat intervals (MFI) in the derivative of the log–log plot. The automatic procedures include the evaluation of the candidate scaling regions using two reliability indices.The data set used to test the methods consists of time series from known model attractors with and without the addition of noise, structured time series, and electrocardiographic signals from the MIT-BIH ECG database.Statistical analysis of results was carried out by means of paired t-test, and no statistically significant differences were found in the large majority of the trials. Consistent results are also obtained dealing with ‘difficult’ time series. In general for a more robust and reliable estimate, the use of both methods may represent a good solution when time series from complex systems are analyzed.Although we present results for the correlation dimension only, the procedures can also be used for the automatic estimation of generalized q-order dimensions and pointwise dimension.We think that the proposed methods, eliminating the need of operator intervention, allow a faster and more objective analysis, thus improving the usefulness of dimension analysis for the characterization of time series obtained from complex dynamical systems. [Copyright &y& Elsevier]