Snu Spaces, from Theory to Practice

Computing the spectrum of singularities of a real-life signal by using the definition is impossible. One rather uses an indirect way to compute it: the multifractal formalism. The first multifractal formalism was introduced by Frisch and Parisi in the context of fully developped turbulence (1985). Its main default is that it always leads to a concave spectrum. For this reason, Stéphane Jaffard has introduced the Snu spaces (2004). They lead to a new multifractal formalism which can detect non concave spectra.
In practice, one has to avoid the concept of limit and to deal with finite size effects. I present a method to determine the spectrum based on the Snu spaces and I illustrate it numerically on theoretical functions.