Wavelet techniques for processing multifractal signals

  • Date: 04/20/2006

Stephane Jaffard (Université de Paris XII)


University of British Columbia


Multifractal analysis has been successfully applied to many types of
signals, starting from turbulence velocity fields, and then extending
to traffic, economic and medecine signals. In each case, it supplies
new parameters which allow for a fine classification of the families of
signals considered. Wavelet methods were introduced in order to obtain
fast and robust numerical algorithms in this field. We will present the
general purposes and techniques of multifractal analysis, and then we
will focus on the recently introduced 'wavelet leaders' method, which
yields the fastest and most accurate numerical methods. Finally, we
will show that it yields an efficient tool to perform classification
and model selection for several types of signals.