This document is organized in six chapters.
Chapter 1 recalls the available methods for processing the recordings of ambient vibrations. An extention to active-source experiments is also detailed. The output of all these techniques is the dispersion curve of surface waves (or a parent curve).
Chapter 2 summarizes all the general options that can be considered to infer the soil properties from an observed dispersion curve. The chosen algorithm (neighbourhood algorithm) is presented with more details. A personal improvement of this technique is discussed at the end.
Chapter 3 presents the algorithm used for computing the dispersion curves for one-dimen-sional models. A number of improvements are proposed to speed up the calculations and to ensure a correct answer. The sensitivity of the dispersion curve to input parameters is tested as well.
Between the inversion algorithm and the forward computation, a crucial step is the parameterization of the ground model. The parameter value ranges are chosen and the prior information is included at this stage. Chapter 4 explains all the strategies for choosing parameter based on synthetic dispersion curve examples.
Chapter 5 details various uncommon inversions, that include higher modes, Love and Ray-leigh modes, the frequency information from H/V techniques, and the direct inversion of auto-correlation curves.
Chapter 6, the array technique is tested on synthetic ambient vibrations with various signal processing methods. The parallel interpretation of arrays of distinct apertures is a key aspect to obtain unbiased dispersion curves, and hence correct 
profiles. The processing techniques are applied to array vibration measurements in the city of Liège, Belgium and the results are compared to other prospecting methods (boreholes, seismic refraction, Cone Penetration Tests, H/V).