next up previous contents
Next: CONCLUSIONS AND SUGGESTIONS Up: APPLICATIONS Previous: Singular values   Contents

Discussion

The dimension estimated using a recording of an alpha rhythm did not saturate; the characteristic rhythm was not even reflected by low-dimensional regions in the correlation integrals as is the case for the respiration. Furthermore, singular value decomposition did not prove to be useful. The spike wave which was oversampled - motivated by the fact that an epileptic seizure is short in duration - showed that we have to be careful about autocorrelations: they may cause a spurious saturation. Irregularities in the respiration signal could not be described by a low-dimensional system. We expect that part of the irregularities are caused by the data acquisition. We have no means, however, to differentiate between data acquisition and biological noise (or chaos). Furthermore, the heart activity, which was visible in the power spectrum of the respiration signal, was not recognizable in the dimension analysis. The confidence intervals in the plots of the estimated dimension and entropies are of little value since we do not know what to do with the estimates themselves7.4.

webmaster@rullf2.xs4all.nl