Wavelets and Time Series Modeling

The aim of this report is to present the most recent ideas related to time-frequency methods and particularily the so called wavelet transform. Time-frequency methods provides with analysis and preprocessing tools in the frequency space opening new perspective for time series modeling. These techniques are expected to provide better analysis tool for nonstationary time series or time series with pseudoperiodic or chaotic behaviour. Moreover, when used to perform preprocessing of the input space, these techniques can enable the use of multi-scale or multi-resolution modeling. Finally, we will examine how they can be put to contribution to address the problem of the complexity determination essential for topological determination and by mean of consequence the generalization issue